Page 1
The Warm–Hot Environment of the Milky Way
DISSERTATION
Presented in Partial Fulfillment of the Requirements for
the Degree Doctor of Philosophy in the
Graduate School of The Ohio State University
By
Rik Jackson Williams
*****
The Ohio State University
2006
Dissertation Committee: Approved by
Professor Smita Mathur, Adviser
Professor David H. Weinberg Adviser
Professor Richard W. PoggeAstronomy Graduate Program
Page 2
ABSTRACT
I present an investigation into the local warm–hot gaseous environment of the
Milky Way as observed through highly ionized metal absorption lines in ultraviolet
and X-ray spectra. These X-ray lines (primarily Ovii) had been reported at redshifts
consistent with zero in previous studies of background quasars; however, it has been
unclear whether this gas exists close to the Galaxy (within a few tens of kpc) or
extends far out into intergalactic space, thereby comprising most of the mass in the
local universe. Additionally, highly–ionized Ovi high–velocity clouds (HVCs), some
of which are associated with the ubiquitous extended neutral hydrogen HVCs seen
around the Galaxy, had been extensively studied. However, the distance to the Ovi
HVCs, and their relation to the X-ray lines, remained undetermined.
With three of the highest–quality Chandra grating spectra of extragalactic
sources to date, a large number of z = 0 absorption lines are detected; the FUSE
spectra of these same objects show low– and high–velocity Ovi absorption. Using
advanced curve–of–growth and ionization balance analysis, limits are placed on the
velocity dispersion, temperature, and density of the warm–hot gas along these lines
of sight. In none of these cases can the absorption be placed conclusively at Galactic
ii
Page 3
or extragalactic distances. However, in two of the three cases (Mrk 421 and Mrk
279), the observed Ovi UV absorption components are found to be inconsistent
with the X-ray absorber, indicating that the X-ray absorption is either extragalactic
or traces a previously undiscovered Galactic component. The third sightline (PKS
2155–304) exhibits absorption with properties more similar to Mrk 421 than Mrk
279; thus, there may be more than one physical process contributing to the observed
absorption along any given sightline.
While the X-ray components of this research exclusively employ Chandra
data, the XMM–Newton mission can in principle be used for the same purpose.
XMM’s effectiveness in observations of WHIM lines is quantitatively analyzed in the
context of two recently detected intervening WHIM systems toward Mrk 421. The
XMM grating spectrograph is found to be inferior to Chandra/LETG due to lower
resolution and narrow detector features that hinder the detection of unresolved lines.
iii
Page 4
Dedicated to Walter J. Williams
iv
Page 5
ACKNOWLEDGMENTS
I cannot thank my advisor, Smita Mathur, enough for the fantastic research
opportunities, constant support, and for fending off the wolves when necessary
(while simultaneously teaching me how to do it for myself). I look forward to many
years of collaboration with her, writing last–minute proposals for observations on
unfamiliar instruments. Many thanks also to Rick Pogge, who (as my effective
first–year advisor) got me started on some excellent projects here at OSU and has
been a continuous source of support and advice throughout. Likewise, I thank
David Weinberg and Andy Gould for their consultation on a number of matters
both political and scientific. None of this would have been possible without Fabrizio
Nicastro and Martin Elvis giving me access to their one-of-a-kind data and teaching
me how to analyze and interpret it.
“I get by with a little help from my friends” (Lennon & McCartney 1967) in
the literal sense, particularly Juna Kollmeier, Reni Ayachitula, Amy Stutz, and Iljie
Kim Fitzgerald. And, of course, a little help from my family, with their unceasing
(if bemused) encouragement and support of my foray into academia.
v
Page 6
Although they may not have been as directly involved in my research as the
people specifically mentioned above, I am indebted to those other past and present
members of the OSU Astronomy Department who have transformed it into the
astronomy field’s foremost venue for graduate research and scientific interaction.
Generous financial support for this work was provided by an Ohio State
University Presidential Fellowship, Chandra award AR5–6017X (issued by the
Chandra X-ray Observatory Center, which is operated for and on behalf of NASA
under contract NAS8–39073), and the National Radio Astronomy Observatory. I
salute the efforts of the Chandra, FUSE, and XMM scientific and support staff for
making these excellent missions possible.
vi
Page 7
VITA
January 29, 1979 . . . . . . . . . . . . . . . Born – Silverton, Oregon, USA
2001 . . . . . . . . . . . . . . . . . . . . . . . . . . . B.S. Astronomy,
California Institute of Technology
2003 . . . . . . . . . . . . . . . . . . . . . . . . . . . M.S. Astronomy, The Ohio State University
2001 – 2002 . . . . . . . . . . . . . . . . . . . . Graduate Fellow, The Ohio State University
2002 – 2005 . . . . . . . . . . . . . . . . . . . . Graduate Research Associate,
The Ohio State University
2005 – 2006 . . . . . . . . . . . . . . . . . . . . . Presidential Fellow, The Ohio State University
PUBLICATIONS
Research Publications
1. I. N. Reid, J. D. Kirkpatrick, J. E. Gizis, C. C. Dahn, D. G. Monet, R.
J. Williams, J. Liebert, and A. J. Burgasser, “Four Nearby L Dwarfs”, AJ, 119, 369,
(2000).
2. J. D. Kirkpatrick, I. N. Reid, J. Liebert, J. E. Gizis, A. J. Burgasser, D.
G. Monet, C. C. Dahn, B. Nelson, and R. J. Williams, “67 Additional L Dwarfs
Discovered by the Two Micron All–Sky Survey”, AJ, 120, 447, (2000).
3. J. E. Gizis, D. G. Monet, I. N. Reid, J. D. Kirkpatrick, J. Liebert, and
R. J. Williams, “New Neighbors from 2MASS: Activity and Kinematics at the
Bottom of the Main Sequence”, AJ, 120, 1085, (2000).
vii
Page 8
4. R. J. Williams, R. W. Pogge, and S. Mathur, “Narrow-Line Seyfert 1
Galaxies from the Sloan Digital Sky Survey Early Data Release”, AJ, 124, 3042,
(2002).
5. S. Mathur, and R. J. Williams, “Chandra Discovery of the Intracluster
Medium Around UM 425 at Redshift 1.47”, ApJ, 589, L1, (2003).
6. R. J. Williams, S. Mathur, and R. W. Pogge, “Chandra Observations of
X-ray Weak Narrow-Line Seyfert 1 Galaxies”, ApJ, 610, 737, (2004).
7. F. Nicastro, S. Mathur, M. Elvis, J. Drake, T. Fang, A. Fruscione, Y.
Krongold, H. Marshall, R. Williams, and A. Zezas, “The mass of the missing baryons
in the X-ray forest of the warm–hot intergalactic medium”, Nature, 433, 495, (3
February 2005).
8. F. Nicastro, S. Mathur, M. Elvis, J. Drake, F. Fiore, T. Fang, A. Frus-
cione, H. Marshall, and R. Williams, “Chandra Detection of Two Warm–Hot IGM
Filaments along the Line of Sight to Mkn 421”, ApJ, 629, 700, (2005).
9. R. J. Williams, S. Mathur, F. Nicastro, M. Elvis, J. J. Drake, T. Fang,
F. Fiore, Y. Krongold, Q. D. Wang, and Y. Yao, “Probing the Local Group Medium
Toward Mkn 421 with Chandra and FUSE”, ApJ, 631, 856, (2005).
10. Q. D. Wang, Y. Yao, T. M. Tripp, T. T. Fang, W. Cui, F. Nicastro, S.
Mathur, R. J. Williams, L. Song, and R. Croft, “Warm–Hot Gas in and around the
Milky Way: Detection and Implications of O VII Absorption Toward LMC X–3”,
ApJ, 635, 386, (2005).
11. R. J. Williams, S. Mathur, F. Nicastro, and M. Elvis, “XMM–Newton
View of the z > 0 Warm–Hot Intergalactic Medium Toward Markarian 421”, ApJ,
642, L95, (2006).
12. R. J. Williams, S. Mathur, and F. Nicastro, “Chandra Detection of
Local Warm–Hot Gas Toward Markarian 279”, ApJ, 645, 179, (2006).
viii
Page 9
FIELDS OF STUDY
Major Field: Astronomy
ix
Page 10
Table of Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Vita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 “Missing Baryons” at Low Redshift . . . . . . . . . . . . . . . . . . . 1
1.2 Relation to Previous Work . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Scope of the Dissertation . . . . . . . . . . . . . . . . . . . . . . . . . 6
Chapter 2 The Markarian 421 Sightline . . . . . . . . . . . . . . . . . . 8
2.1 Observations and Data Preparation . . . . . . . . . . . . . . . . . . . 8
2.1.1 Chandra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 FUSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Line Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Absorption Line Diagnostics . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Doppler Parameters . . . . . . . . . . . . . . . . . . . . . . . . 15
x
Page 11
2.3.2 Column Densities . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.3 Temperature and Density Constraints . . . . . . . . . . . . . . 19
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.1 Potential Caveats . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4.2 Where does the X-ray absorption originate? . . . . . . . . . . 30
2.4.3 Comparisons to Other Studies . . . . . . . . . . . . . . . . . . 33
2.5 Summary and Future Work . . . . . . . . . . . . . . . . . . . . . . . 34
Chapter 3 The Markarian 279 Sightline . . . . . . . . . . . . . . . . . . 50
3.1 Data Reduction and Measurements . . . . . . . . . . . . . . . . . . . 50
3.1.1 Chandra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.1.2 FUSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2.1 Doppler Parameters and Column Densities . . . . . . . . . . 56
3.2.2 Temperature and Density Diagnostics . . . . . . . . . . . . . . 59
3.2.3 The AGN Warm Absorber . . . . . . . . . . . . . . . . . . . . 63
3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3.1 Comparison to the Mrk 421 Sightline . . . . . . . . . . . . . . 64
3.3.2 Origin of the Absorption . . . . . . . . . . . . . . . . . . . . . 64
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Chapter 4 The PKS 2155–304 Sightline . . . . . . . . . . . . . . . . . . 77
4.1 Data Reduction and Measurements . . . . . . . . . . . . . . . . . . . 78
4.1.1 Chandra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
xi
Page 12
4.1.2 FUSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2.1 Doppler Parameters and Column Densities . . . . . . . . . . 84
4.2.2 Temperature and Density Diagnostics . . . . . . . . . . . . . 87
4.2.3 z = 0.055 Absorption Reported by Fang et al. . . . . . . . . . 91
4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.3.1 Comparison to Other Lines of Sight . . . . . . . . . . . . . . . 92
4.3.2 Where is the Absorption? . . . . . . . . . . . . . . . . . . . . 96
4.3.3 Comparison to Nicastro et al. (2002) . . . . . . . . . . . . . . 97
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Chapter 5 Instrumental Considerations: Chandra or XMM–Newton? . 114
5.1 Data Reduction and Measurements . . . . . . . . . . . . . . . . . . . 115
5.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.3 Disputed Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Chapter 6 Summary and Future Work . . . . . . . . . . . . . . . . . . 128
6.1 Individual X-ray Sightlines . . . . . . . . . . . . . . . . . . . . . . . . 128
6.2 The Importance of Spectral Fidelity . . . . . . . . . . . . . . . . . . . 130
6.3 Future Prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.3.1 X-ray Observations . . . . . . . . . . . . . . . . . . . . . . . . 131
6.3.2 Longer Wavelengths . . . . . . . . . . . . . . . . . . . . . . . 133
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
xii
Page 13
List of Tables
2.1 Observed z ≈ 0 lines. . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.1 Observed z ≈ 0 lines. . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1 Observed z ≈ 0 absorption lines . . . . . . . . . . . . . . . . . . . . . 69
4.1 Chandra observation log . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.2 Observed z ≈ 0 absorption lines . . . . . . . . . . . . . . . . . . . . . 101
4.2 Observed z ≈ 0 absorption lines . . . . . . . . . . . . . . . . . . . . . 102
5.1 XMM–Newton observation log . . . . . . . . . . . . . . . . . . . . . 123
5.2 Absorption line equivalent width measurements . . . . . . . . . . . . 124
xiii
Page 14
List of Figures
2.1 Mkn 421 Chandra LETG spectrum . . . . . . . . . . . . . . . . . . . 40
2.2 Mrk 421 FUSE spectrum near the Ovi line . . . . . . . . . . . . . . . 41
2.3 Ovii curve–of–growth diagnostics . . . . . . . . . . . . . . . . . . . . 42
2.4 Ovi curve–of–growth diagnostics . . . . . . . . . . . . . . . . . . . . 43
2.5 Temperature and density diagnostics from oxygen lines . . . . . . . . 44
2.6 Temperature and density diagnostics with solar abundances . . . . . 45
2.7 Temperature and density diagnostics with shifted abundances . . . . 46
2.8 Ionic abundance models for the cooler (likely Galactic) ions . . . . . . 47
2.9 Ionic abundances vs. temperature for possible extragalactic ions, low
density case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.10 Ionic abundances vs. temperature for possible extragalactic ions, high
density case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Full Chandra spectrum of Mrk 279 . . . . . . . . . . . . . . . . . . . 70
3.2 18–22 A Chandra spectrum of Mrk 279 . . . . . . . . . . . . . . . . . 71
3.3 Velocity plots of the local Ovii and Ovi absorption lines . . . . . . . 72
3.4 Curve–of–growth diagnostics for the Ovii K–series . . . . . . . . . . 73
3.5 Curve–of–growth analysis for the Ovi UV absorption . . . . . . . . . 74
3.6 Temperature and density constraints from Ovii and Ovi, b = 100 km s−1 75
3.7 Temperature and density constraints, b = 200 km s−1 . . . . . . . . . 76
xiv
Page 15
4.1 Chandra ACIS–S/LETG continuum fit . . . . . . . . . . . . . . . . . 103
4.2 Chandra HRC–S/LETG continuum fit . . . . . . . . . . . . . . . . . 104
4.3 Detected z = 0 absorption lines (ACIS–S/LETG) . . . . . . . . . . . 105
4.4 Detected z = 0 absorption lines (HRC–S/LETG) . . . . . . . . . . . 106
4.5 1032 Aregion of the FUSE spectrum . . . . . . . . . . . . . . . . . . . 107
4.6 Ovii curve–of–growth analysis . . . . . . . . . . . . . . . . . . . . . . 108
4.7 Ovi curve–of–growth analysis . . . . . . . . . . . . . . . . . . . . . . 109
4.8 Oxygen ion temperature and density constraints . . . . . . . . . . . . 110
4.9 X-ray ion temperature and density constraints (low–b) . . . . . . . . 111
4.10 X-ray ion temperature and density constraints (high–b) . . . . . . . . 112
4.11 Chandra spectrum near the Oviii z = 0.055 wavelength . . . . . . . . 113
5.1 XMM–Newton spectrum of Mrk 421 . . . . . . . . . . . . . . . . . . . 125
5.2 RGS1 and RGS2 instrumental response functions . . . . . . . . . . . 126
5.3 Line–spread functions for Chandra and XMM–RGS . . . . . . . . . . 127
xv
Page 16
Chapter 1
Introduction
“Look on my works, ye mighty, and despair!”
— Ozymandias, Percy Bysshe Shelley
Gonna get my PhD
I’m a teenage lobotomy
— Teenage Lobotomy, The Ramones
1.1. “Missing Baryons” at Low Redshift
Over the past fourteen billion years, the baryonic mass found in the intergalactic
medium (IGM)–a tenuous web of gas bridging the gaps between galaxies and clusters–
is thought to outweigh the baryons found in all other sources, stars, galaxies, and
the hot gas that dominates the mass of clusters of galaxies. Indeed, at high redshifts
(z ∼> 2) the “forest” of Lyα absorption lines seen in spectra of distant quasars reveals
a vast network of cool, photoionized hydrogen that is consistent with the expected
baryon density at those redshifts (Weinberg et al. 1997). At more recent epochs,
1
Page 17
however, the process of structure formation has shock–heated this intergalactic gas
to produce the warm–hot IGM (WHIM; Cen & Ostriker 1999; Dave et al. 2001) with
temperatures of T ≈ 105 − 107 K densities 10−6 − 10−4 cm−3, or cosmic overdensities
of δ ≈ 10 − 100.
This WHIM gas has proved extremely difficult to detect, resulting in a
discrepancy between the observed baryon census and predictions from the cosmic
microwave background (Bennett et al. 2003). At such low densities and high
temperatures, the combination of collisional– and photoionization renders most of
the gas too highly ionized to be detected through its Lyα absorption, though some
broad Lyα systems at low redshift have been reported (Sembach et al. 2004; Richter
et al. 2004). Moreover, its extremely low density prevents thermal and/or line
emission from the WHIM from being detected even with the most sensitive current
instruments. Heavier elements such as oxygen, nitrogen, and neon would be highly
(but not fully) ionized in such a medium, and these metals are predicted to provide
a unique view of the WHIM through their higher–energy UV and X-ray resonance
absorption lines (Perna & Loeb 1998; Hellsten et al. 1998; Fang, Bryan, & Canizares
2002).
Though they are quite weak, these WHIM lines are now detectable in principle
with the advent of such facilities as the Chandra X-ray Observatory, XMM–Newton,
the Hubble Space Telescope, and the Far Ultraviolet Spectroscopic Explorer (FUSE).
Detections of such intervening WHIM filaments, with varying levels of confidence,
2
Page 18
have been reported along several lines of sight (Nicastro et al. 2005a,b; Mathur et al.
2003; Fang et al. 2002). The total baryonic mass reported by Nicastro et al. (2005a)
is indeed consistent (within the admittedly large errors) with the aforementioned
baryon deficit at low redshifts.
Since most galaxies are expected to trace the same cosmic overdensities as the
web of WHIM filaments, it would be no surprise if the Galaxy itself resided in such
a filament. Indeed, X-ray spectra of several quasars have shown likely z = 0 Ovii
absorption, but it is unclear whether this absorption is actually due to the nearby
WHIM or is instead a component of the Galaxy itself, such as a hot halo or corona
(or some combination of the two). Some Ovii absorption has indeed been found
within 50 kpc of the Galaxy (Wang et al. 2005), but this is unlikely to be uniformly
distributed. Simulations of the Local Group strongly indicate that a large amount
of warm–hot gas is expected near zero redshift (Kravtsov et al. 2002). Thus, in
reality the X-ray absorption is likely to be caused by a combination of Galactic
and extragalactic components, perhaps with one dominating the other in certain
directions.
Further complicating the issue is the presence of other gaseous components
of unknown origin. H I high–velocity clouds (HVCs) have a velocity distribution
inconsistent with Galactic rotation and therefore are thought to be either neutral
gas high in the Galactic halo or cooling, infalling gas from the surrounding IGM.
Along many lines of sight studied with FUSE, high–velocity Ovi absorption lines at
3
Page 19
velocities coincident with the H i HVCs are seen, while in some other directions the
high–velocity Ovi is present even in the absence of H i emission at that velocity
(Sembach et al. 2003). Some of these latter, unassociated Ovi HVCs were found to
be at rest in the Local Group rest frame (as a population), indicating that they may
be extragalactic in origin (Nicastro et al. 2003). On the other hand, many of these
Ovi HVCs also show absorption from lower ionization states that are unlikely to
arise in a low–density, warm–hot IGM (Sembach 2003).
While the evidence appears to point to both Galactic and extragalactic
characteristics for the Ovi HVCs, their connection to the highly–ionized gas seen
in X-rays (if any) is unknown. Part of the problem has been the tremendous
amount of Chandra observing time that is required to obtain a high–quality grating
spectrum of an extragalactic source: since X-ray telescopes are essentially photon
counting devices, thousands of scarce (compared to optical telescopes), high–energy
X-ray photons must be collected for each resolution element in order to detect even
relatively strong WHIM lines (∼ 20 mA, corresponding to NOVII ≈ 1016 cm−2). In
the past few years, however, there have been several opportunities to overcome this
difficulty: observing AGN only during extremely bright flares (as with Mrk 421),
re–analyzing long exposures that were originally performed for other purposes (Mrk
279), and co–adding many short calibration observations of the same object taken
over the past seven years (PKS 2155–304).
4
Page 20
The three aforementioned AGN have thus been observed with these techniques
to unprecedented levels of sensitivity with the Chandra gratings. In this dissertation
I present measurements and analyses of the z ≈ 0 X-ray absorption lines seen in
these spectra with the goal of determining the location and physical properties of the
absorbing material, and its connection to the Ovi absorption seen along each line of
sight in FUSE. Such high quality X-ray data have not been previously available, and
new techniques are developed to obtain the greatest amount of information on the
physical state of the absorbing medium (while overcoming some unique aspects of
the data, such as a complicated instrumental response function and far lower spectral
resolution than is available in the FUSE data). The efficacy of the XMM–Newton
observatory for these studies is also investigated.
1.2. Relation to Previous Work
Prior to this work, a few studies had reported the presence of z = 0 Ovii
absorption, namely toward 3C 273 (Fang et al. 2003) and PKS 2155–304 (Nicastro
et al. 2002). Since this latter detection was published, far more X-ray data became
available in the form of calibration observations; the analysis in Chapter 4 takes all
this available data into account and compares my new results to those of Nicastro
et al. (2002). New observations of 3C 273 are also available; however, hot gas
from a supernova remnant is likely to lie along this line of sight. Due to the low
5
Page 21
velocity resolution of Chandra (∼ 700 km s−1 at the Ovii wavelength), absorption
from the remnant would be fully blended with any WHIM or Galactic corona Ovii,
making this sightline of limited use. Constrained simulations of the Local Group by
Kravtsov et al. (2002) provide a strong theoretical basis for the existence of z = 0
WHIM while predicting roughly where the strongest absorption can be expected.
In fact, far more literature on the z > 0 WHIM has been published. Fang et al.
(2002), Mathur et al. (2003), and McKernan et al. (2003) reported early tentative
detections of the WHIM toward several sources, and I address the Fang et al. (2002)
detection of z = 0.055 Oviii toward PKS 2155–304 in Chapter 4. By far the most
confident detection of the z > 0 WHIM to date is in the Mrk 421 spectrum (which
I analyze in Chapter 2) by Nicastro et al. (2005a). Two WHIM filaments were
unambiguously detected in this spectrum, allowing the temperature density, and
relative metal abundances of the WHIM to be estimated. These provide a valuable
set of parameters to compare with those derived from the z = 0 absorption. The
detections of these WHIM filaments are also revisited with XMM–Newton in Chapter
5.
1.3. Scope of the Dissertation
The data, analysis, and interpretation of the X-ray and UV data along three
extragalactic sightlines are the focus of this dissertation. Owing to the individual
6
Page 22
peculiarities of the observations and the variety of analysis techniques required
for the different data, the three following chapters are each devoted to one line of
sight: Markarian 421 (Chapter 2), Markarian 279 (Chapter 3), and PKS 2155–304
(Chapter 4). Chapter 5 presents a comparison of the XMM–Newton data of Mrk
421 with the previously reported Chandra detection of two WHIM filaments along
this line of sight, and quantitatively describes why XMM is unable to detect these
filaments. Finally, I summarize the results of this dissertation and comment on
present and future avenues for this research.
A major fraction of the research presented in this dissertation has been
published in the scientific literature. Chapter 2 is largely taken from Williams et
al. 2005, ApJ, 631, 856; chapter 3 has appeared as Williams et al. 2006, ApJ, 645,
179; and most of chapter 5 also appears in Williams et al. 2006, ApJ, 642, L95.
7
Page 23
Chapter 2
The Markarian 421 Sightline
Through a program of observing nearby blazars in outburst phases, we have
obtained high–quality Chandra and FUSE spectra of Mkn 421, sufficient to study in
detail the local WHIM (and Galactic halo/thick disk) absorption. Here I report on
these observations, and the inferred properties of the local absorption.
2.1. Observations and Data Preparation
2.1.1. Chandra
A full description of the Chandra observations, data reduction, and continuum
fitting can be found in Nicastro et al. (2005a); a brief summary follows. Mkn 421
was observed during two exceptionally high outburst phases for 100 ks each as part
of our Chandra–AO4 observing program: one at f0.5−2keV = 1.2 × 10−9 erg s−1 cm−2
with the Low Energy Transmission Grating (LETG) combined with the Advanced
CCD Imaging Spectrometer–Spectroscopic (ACIS-S; Garmire et al. 2003) array,
and another at f0.5−2keV = 0.8 × 10−9 erg s−1 cm−2 with the High Resolution
8
Page 24
Camera–Spectroscopic (HRC-S; Murray & Chappell 1985) array and LETG. Each
of these observations contains ∼ 2500 counts per resolution element at 21.6 A.
Additionally, another short observation of Mkn 421 was taken with HRC/LETG
(29 May 2004), providing another 170 counts per resolution element. These three
spectra were combined over the 10–60 A range to improve the signal–to–noise ratio
(S/N≈ 55 at 21 A with 0.0125 A binning). The final coadded spectrum of Mkn 421
is one of the best ever taken with Chandra: it contains over 106 total counts with
∼ 6000 counts per resolution element at 21.6 A, providing a 3σ detection threshold
of Wλ ≈ 2 mA (NOVII = 8 × 1014cm−2 for an unsaturated line).
Effective area files (ARFs) for each observation were built using CIAO1 v3.0.2
and CALDB2 v2.2.6. Those pertaining to the ACIS/LETG observations were
corrected for the ACIS quantum efficiency degradation3 (Marshall et al. 2003). For
the HRC/LETG observations, the standard ARFs were used. Each ARF was then
convolved with the relevant standard redistribution matrix file (RMF), and the
convolved RMFs were weighted by exposure time, rebinned to the same energy scale,
and averaged to provide a response file for the coadded spectrum.
Using the CIAO fitting package Sherpa4, we initially modeled the continuum
of Mkn 421 as a simple power law and a Galactic absorbing column density of
1cxc.harvard.edu/ciao/
2cxc.harvard.edu/caldb/
3See also cxc.harvard.edu/ciao/why/acisqedeg.html
4cxc.harvard.edu/sherpa/
9
Page 25
NH = 1.4 × 1020 cm−2 (Dickey & Lockman 1990), excluding the 48–57 A HRC chip
gap region. Metal abundances for the Galactic gas were then artificially adjusted to
provide a better fit around the O I and C I K–edges near 23 A and 43 A respectively.
This is not intended to represent actual changes to the absorber composition, but
rather to correct uncertainties in the instrument calibration. These adjustments
affect the continuum mostly near the carbon, oxygen, and neon edges, but individual
narrow absorption lines are unaffected. After this fit there were still some systematic
uncertainties in the best–fit continuum model; these were corrected with broad
(FWHM = 0.15 − 5 A) Gaussian emission and absorption components until the
modeled continuum appeared to match the data upon inspection. Indeed, the
residuals of the spectrum to the final continuum model have a nearly Gaussian
distribution, with a negative tail indicating the presence of narrow absorption lines
(see Nicastro et al. 2005a, Figure 8).
2.1.2. FUSE
Mkn 421 was also observed with FUSE as part of our Director’s Discretionary
Time observing program on 19–21 January 2003 for a total of 62.8 ks. An additional
21.8 ks observation from 1 December 2000 was also available in the archive. We used
the time–tagged, calibrated data from only the LiF1A detector, since inclusion of
the LiF2B data provides a small (∼ 20%) increase in S/N but degrades the overall
10
Page 26
spectral resolution.5 These two observing programs consist of four observations,
which in turn contain a total of 29 individual exposures. The wavelength scales of
each observation’s constituent exposures were cross–correlated and shifted (typically
by 1–2 pixels) to account for slight uncertainties in the wavelength calibration. The
exposures for each observation were checked for consistency and coadded, weighted
by exposure time. The resulting four spectra were then cross-correlated against
each other, coadded (with a ∼ 10% downward shift in flux applied to the 2000
observation due to source variability), and binned by 5 pixels (0.034 A, or one–half
of the nominal 20km s−1 resolution) providing a S/N of 17 near 1032 A.
To check the absolute wavelength calibration we followed the method of Wakker
et al. (2003), using their 4–component fit to the Murphy et al. (1996) Green Bank
H I–21 cm data as a velocity reference. They find four main components of H I
emission with an NH–weighted average velocity of −31.7km s−1. In the FUSE
spectrum, the Si II λ1020.699 A and Ar I λ1048.220 A lines are expected to trace the
same gas as the H I emission. Each UV line was fit with two Gaussian components
in Sherpa, giving average velocity offsets of −30.9 and −34.9km s−1 respectively.
These agree well with the H I data, though the slight difference between the Ar I and
Si II measurements suggest at least a ∼ 4km s−1 intrinsic wavelength uncertainty.
5See the FUSE Data Analysis Cookbook v1.0, fuse.pha.jhu.edu/analysis/analysis.html
11
Page 27
2.2. Line Measurements
To find and identify narrow absorption lines in the Chandra spectrum of
Mkn 421, we visually inspected small (2–5 A) regions of the spectrum, beginning
around the rest wavelength of OviiKα (21.602 A) since this tends to be the strongest
z = 0 X-ray absorption line (e.g. Nicastro et al. 2002; Fang et al. 2003; Chen et al.
2003). Three OviiKα (Figure 2.1) absorption features were found: one at z = 0,
one at z = 0.011, and one at z = 0.027 (with typical redshift errors of 0.001).
There is also a strong feature which is ∼ 3σ from the OvKα rest wavelength, but is
more likely OviiKα at v ≈ +900km s−1 relative to the blazar. A close pair of lines
consistent with Lyα at this velocity has been observed (Shull et al. 1996; Penton et
al. 2000), so this may be indicative of an inflow to Mkn 421 or uncertainty in the
blazar redshift (based on rather old spectrophotometric measurements by Margon
et al. 1978). A weak OviKα line is seen at the rest wavelength of 22.02 A. Other
regions of the spectrum were then searched for lines corresponding to these systems,
with particular emphasis paid to strong transitions of the most abundant elements
(C, N, O, and Ne). All in all there were 13 lines marginally or strongly detected
at z ≈ 0 (including the Nvii, Ov, and Arxv upper limits), 3 at z = 0.011, and 7
at z = 0.027. The latter two systems are the subject of other papers (Nicastro et
al. 2004a,b) and thus will not be discussed further here.
12
Page 28
These 13 z = 0 X-ray lines were fitted in Sherpa with narrow Gaussian features
superposed on the fitted continuum described in §2.1.1 (see Figure 2.1). We are
excluding the strong O I (23.51 A) line since it arises in the neutral ISM and is
not of interest here, as well as the O2 (23.34 A) absorption since it coincides with
a strong instrumental feature and cannot be accurately measured. Due to the
FWHM = 0.04 A (∼ 600km s−1) LETG resolution the lines are all unresolved,
so only the position and equivalent width of each line are measured. Errors are
calculated using the “projection” command in Sherpa, allowing the overall continuum
normalization to vary along with all parameters for each line. The resulting line
parameter estimates are presented in Table 2.1. The ∼ 0.02 A systematic wavelength
uncertainty of the LETG6 is in most cases larger than the statistical uncertainty
of the line centroid; thus, Table 2.1 lists whichever is greater. Additionally, a
meaningful upper error bar on the Cvi equivalent width could not be calculated
with Sherpa. In this case, the FWHM was frozen at the instrumental resolution and
the error was recalculated; a visual inspection confirms the new limit to be more
reasonable. Upper limits for the Ov, Nvii, Arxv, and Nex lines were calculated
with both the position and FWHM frozen.
The FUSE spectrum (Figure 2.2) shows a strong, broad low–velocity Ovi
1032 A absorption line at z ≈ 0 due to gas in the Galactic thick disk and halo
(Savage et al. 2003). An asymmetric wing on the red side of this line is evident,
6cxc.harvard.edu/cal/
13
Page 29
possibly a kinematically distinct HVC. We fitted the Ovi 1032 A line in Sherpa
using a constant local continuum (in a ±2 A window) and two Gaussian absorption
components: one for the v ≈ 0 OviLV line, and one at v ≈ 100km s−1 for the
HVC. No H2 contamination is seen at the Ovi 1032 A wavelength when absorption
templates are fit to the other H2 lines seen in the spectrum. The 1037 A line is
somewhat blended with a single H2 absorption line; this is taken into account with
another narrow Gaussian. From this fit, we find equivalent widths of 18.6 ± 5.6 mA
for the 1032 A HVC and 270.7 ± 7.9 mA for the Galactic component. The best–fit
model for the HVC is fairly robust and not sensitive to variations in the initial
parameters; however, the derived equivalent width of 18.5 ± 5.6 mA is lower than
the 37 ± 11 ± 29 mA (errors are statistical and systematic, respectively) measured
by Wakker et al. (2003) in the initial 21.8 ks observation. They employed a direct
integration method which may not have taken into account the substantial blending
of the Galactic OviLV with the HVC. Our total Ovi equivalent width (LV+HVC
= 279 ± 10 mA) is in good agreement with their value of 285 ± 20 mA.
Deblending the Ovi 1037 A line is less certain due to the presence of adjacent
Galactic C II and H2 absorption. A flat continuum was again employed from
1035 − 1040 A and single Gaussian components were used to fit the C II, Ovi,
and H2 absorption lines. The HVC on the 1032 A Ovi line should also appear at
∼ 1038 A with Wλ = 0.50×Wλ(1032 A). Although this component is too weak to be
detected directly, it could cause the measurement of the Galactic LV–Ovi 1037 A
14
Page 30
line to be systematically high. Another absorption Gaussian with one–half of the
1032 A HVC equivalent width (and with the same FWHM and velocity offset) was
included in the 1037 A Ovi line fit to account for this. Table 2.1 lists the measured
properties of the OviLV and HVC absorption lines.
2.3. Absorption Line Diagnostics
2.3.1. Doppler Parameters
To convert the measured equivalent widths to ionic column densities, we
calculated curves of growth for each absorption line over a grid of Doppler
parameters (b = 10− 100km s−1) and column densities (log NH/cm−2 = 12.0− 18.0),
assuming a Voigt line profile. Since the X-ray lines are unresolved, b cannot be
measured directly. It can, however, be inferred from the relative strengths of the three
measured Ovii K–series lines. These lines are produced by the same ionic species,
so in an unsaturated medium Wλ ∝ fluλ2 where flu is the oscillator strength. The
expected equivalent width ratio of Ovii Kβ to Kα is then Wλ(Kβ)/Wλ(Kα) = 0.156,
so the measured value of 0.49 ± 0.09 indicates that the Kα line is saturated. On the
other hand, the measured Ovii Kγ/Kβ ratio is 0.43 ± 0.16, in agreement with the
predicted (unsaturated) value of 0.34.
15
Page 31
These line ratios by themselves are insufficient to determine the physical state
of the Ovii–absorbing medium since b and NOVII are degenerate: the Kα line
saturation could be due to high column density, low b, or a combination of both.
However, given an absorption line with a measured equivalent width and known
fluλ2 value, the inferred column density as a function of the Doppler parameter can
be calculated. The measured equivalent width (and errors) for each transition thus
defines a region in the NOVII − b plane. Since the actual value of NOVII is fixed, b and
NOVII can be determined by the region over which the contours “overlap;” i.e. the
range of Doppler parameters for which the different transitions provide consistent
NOVII measurements.
Figure 2.3 shows such 1σ contours for the three measured Ovii transitions.
As expected, the inferred NOVII is nearly constant in the unsaturated regime (large
b), and rises sharply at low b as the lines begin to saturate. At each value of b, the
differences ∆(log Nαβ) = log(NKα)−log(NKβ) and ∆(log Nαγ) = log(NKα)−log(NKγ)
were calculated, along with the errors on each ∆(log N). The quantity ∆(log Nαβ)
is consistent with zero at the 1σ and 2σ levels for 15 < b < 46km s−1
and 13 < b < 55km s−1 respectively, while ∆(log Nαγ) provides limits of
31 < b < 50km s−1 and 24 < b < 76km s−1 respectively. Since ∆(log Nαγ) provides a
more stringent lower limit on b while ∆(log Nαβ) better constrains the upper limit, we
thus assume a 1σ range of 31 < b < 46km s−1, and a 2σ range of 24 < b < 55km s−1.
It should be noted that Figure 2.3 also shows some overlap between the Kα and Kγ
16
Page 32
at b ∼< 12km s−1; however, this solution is unlikely given the lower limit provided by
the Kβ line. Moreover, b = 12km s−1 implies a maximum temperature (assuming
purely thermal motion) of Tmax = 1.3 × 105 K; such a low temperature is unlikely to
produce the observed strong high–ionization lines.
A similar analysis is not as effective when applied to the strong OviLV UV
doublet (from the thick disk), since these lines are only slightly saturated: the
measured Wλ ratio is 0.61 ± 0.04, compared to the expected unsaturated value of
0.50. When the inferred NOVI is calculated as a function of b for both lines of the
OviLV doublet, the predicted NOVI values are consistent over b = 34 − 112km s−1
(at the 2σ level; see Figure 2.4). Since the OviLV 1032 A line is fully resolved by
FUSE (∼ 15 resolution elements across the line profile) and relatively unblended,
its Doppler parameter can be estimated much more accurately using the measured
line width and strength. In an unsaturated absorption line, FWHM = 2(ln 2)1/2b;
however, the measured FWHM increases if the line is saturated. We compensated for
this by calculating Voigt profile FWHMs on a grid of NOVI and b, and determining
the region consistent with the OviLV 1032 A FWHM measurement of 152± 7km s−1
(or b = 91 ± 4km s−1 assuming no saturation).
When the FWHM–derived contour is overlaid on the NOVI − b contour
inferred from the equivalent width measurement of the LV–Ovi 1032 A line, the
two regions overlap nearly orthogonally (Figure 2.4) leading to a constraint of
b(OviLV)= 80.6± 4.2km s−1. This is ∼ 2σ lower than the unsaturated FWHM, once
17
Page 33
again confirming that the OviLV is only weakly saturated. At this b the inferred
column densities from the two lines of the OviLV doublet differ by 1.3σ but this is
only a minor discrepancy and likely due to errors introduced by the blending of the
1037 A line; thus, we will only consider results from the more reliable 1032 A line
measurement. However, at no value of the Doppler parameter do the 1032 A, 1037 A,
and OviKα lines all produce a consistent NOVI measurement; in fact, the OviKα
column density is a factor of ∼ 4 higher than that inferred from the UV data. This
discrepancy is discussed further in §2.4.1.
2.3.2. Column Densities
The Doppler parameters measured for the Ovii (31km s−1 < b < 46km s−1)
and OviLV (b = 80.6 ± 4.2km s−1) absorption are inconsistent at the ∼ 3σ level,
indicating the presence of at least two distinct components: the Galactic thick–disk
gas traced by broad v ≈ 0 OviLV absorption, and another lower–b phase, possibly
of extragalactic origin, traced by the Ovii absorption lines. It cannot be assumed
a priori that any given line (other than those used to determine b) originates in
one phase or another; moreover, the uncertainty in the calculated column density
depends not only on the equivalent width error but also the error in b. To take
this into account, for each ion the derived column density log Ni and its ±1σ limits
were averaged over the ±1σ ranges of both measured Doppler parameters. As it
turns out, the choice of b does not make a significant difference since all other
18
Page 34
lines (besides the Ovii and OviLV absorption) are essentially unsaturated; i.e., the
difference in Ni calculated with the OviLV and Ovii Doppler parameters is small
compared to the 1σ error on the equivalent width measurements. Even so, to avoid
possible systematic errors, we assumed b = 80.6 ± 4.2km s−1 for those lines likely to
originate in the Galactic thick disk (OviLV, Ov, and Cv), and b = 31 − 46km s−1
for all other species. The derived ionic column densities are listed in Table 2.1
2.3.3. Temperature and Density Constraints
At densities such as those found in the Galactic interstellar medium (ISM;
ne ≈ 1 cm−3), photoionization is unimportant because thermal collisions are by far
the dominant ionization source. This is also the case for very high temperatures
(T ∼> 107 K) even at low densities, since the collisional rate is greater than
the photoionization rate. However, at the low densities typically found in the
intergalactic medium (ne = 10−6 − 10−4 cm−3), photoionization from the diffuse
UV/X-ray background begins to play a greater role by enriching the abundances of
highly–ionized elements at typical WHIM temperatures (log T (K) ≈ 5 − 7) relative
to those expected from pure collisional ionization (Nicastro et al. 2002; Mathur et al.
2003). It is thus imperative that the ionizing background be taken into account in
order to accurately predict ionic abundances in the WHIM.
19
Page 35
Version 90.04 of the ionization balance code Cloudy (Ferland 1996) was used to
compute collisional– plus photoionization hybrid models for the absorbing medium.
Relative ionic abundances were computed over a grid of log T (K) = 4.5 − 7.4 and
log ne(cm−3) = −7−0, with a step size of 0.1 dex in both log ne and log T . Initially, a
rigid scaling of [Z/H] = −1 for all metals was assumed. For the ionizing background
we employed the Sternberg et al. (2002) fit to the metagalactic radiation field:
Jν =
Jν0
(
νν0
)
−3.131 < ν
ν0
< 4
2.512 × 10−2Jν0
(
νν0
)
−0.46νν0
> 4
(2.1)
where here Jν0 = 2×10−23 ergs s−1 cm−2 Hz−1 sr−1 and ν0 = 13.6 eV. The total flux of
ionizing photons is then given by fγ = 4π∫
∞
ν0(Jν/hν)dν = 1.3×104 photons s−1 cm−2,
and the ionization parameter is log U = log(fγ/c) − log ne = −6.36 − log(ne) where
ne is the electron density in cm−3.
Using the ionic abundances calculated with Cloudy, we derived expected
abundance ratios for all observed ions at each point in the log ne − log T plane.
Since any given density and temperature uniquely determines a set of abundance
ratios (Na/Nb for all ions a and b), the problem can be inverted: any value of Na/Nb
defines a curve in the log ne − log T plane, i.e. a set of temperatures and densities
which can produce the measured ratio. When the errors on Na/Nb are taken into
account, the curves become contours, and the overlap between two or more contours
defines the temperatures and densities for which the measured ratios are consistent.
20
Page 36
This is analogous to the method used in §2.3.1 to determine Doppler parameters for
OviLV and Ovii.
The most powerful diagnostics are those using ratios between different ions of
the same element, since these ratios are independent of the relative metal abundances.
Unfortunately the Nvii/Nvi and Nex/Ne ix upper limits are not stringent enough
to place meaningful constraints on the temperature and density. Figure 2.5 shows
the log ne − log T contours for ratios between the X-ray OviKα, Ovii, and Oviii
lines as well as the OviHVC/Ovii ratio. The X-ray line ratios are inconsistent with a
high–density (ne ∼> 10−3cm−3), high–temperature (log T > 6.2) medium, and instead
converge on a partially photoionized plasma with ne = 10−4.7 − 10−3.9 cm−3 (from
the overlap between the OviKα/Ovii and Oviii/Ovii contours) and T = 105.5−5.7 K
(from the limits provided by OviKα/Ovii in this density range). These ranges of
temperatures and densities are in line with those expected from WHIM gas (Dave et
al. 2001). Of course, this is all contingent on the OviKα line being a reliable tracer
of NOVI; this caveat is discussed in detail in §2.4.
On the other hand, the OviHVC/Ovii ratio overpredicts the temperature by
at least an order of magnitude for all values of log ne—in order to be consistent
with the Oviii/Ovii ratio, the OviHVC/Ovii ratio would need to be stronger by a
factor of ∼ 2.5 (or ∼ 3.5σ). It is possible that the HVC is not a physically distinct
component, but is instead the result of some systematic error (such as fixed pattern
noise or an unexpected anomaly in the Galactic OviLV velocity distribution). In this
21
Page 37
case, the Ovi associated with the Ovii and Oviii may be completely blended with
the thick–disk OviLV and thus unmeasurable. Consistency with the Oviii/Ovii
ratio (in the collisional ionization regime) requires log(NOVI/NOVII) ≈ −2.5, or
roughly 20% of the Galactic UV Ovi absorption. Although it appears that the Ovi
HVC as measured cannot originate in the same medium as the Ovii absorption, we
suspect that additional atomic physics may be at work here and could in principle
reconcile this disagreement (see §2.4.1).
While the OviKα/Ovii and Oviii/Ovii ratios provide strong constraints,
it is important to consider other ion ratios as well (particularly since the OviKα
and 1032 A Ovi column densities disagree). Figure 2.6 shows the log ne − log T
contours for several different ion ratios, all calculated relative to Ovii since the error
on NOVII is small. With a rigid metallicity shift relative to solar, the Ne ix/Ovii
Oviii/Ovii, and OviKα/Ovii ratios are not all consistent with each other for
any combination of temperature and density; however, the consistency can be
improved with adjustments to the [Ne/O] ratio (see §2.4.1). Both the Cvi/Ovii
and Nvii/Ovii measurements are consistent with a high– or low–density medium
at solar abundances.
Limits on the temperature of the Galactic thick–disk absorption can be derived
in a similar fashion, although it is not the primary focus of this work and there are
far fewer measured lines to work with. The most accurately–measured line is the
OviLV; additionally, Cv and Nvi X-ray lines are measured, and upper limits have
22
Page 38
been determined for Ov and Nvii. Figure 2.8 shows the temperature constraints
derived for this Galactic absorption, assuming pure collisional ionization. The
Ov/OviLV and Nvii/Nvi upper limits provide metallicity–independent constraints
of log T > 5.39 and log T < 6.64 respectively. A more stringent upper limit on
temperature of log T < 6.03 is provided by the Nvii/OviLV ratio, but this is
somewhat dependent on [N/O]. Within this range the Cv/OviLV ratio provides an
even stricter limit of 5.3 < log T < 5.7, but again this depends on [C/O]. At these
temperatures the expected Ovii column density is at most an order of magnitude
less than measured; thus, the OviLV, Ovii, and Oviii cannot all originate in the
same phase assuming pure collisional ionization (see also Mathur et al. 2003).
2.4. Discussion
Our Chandra and FUSE observations have provided a wealth of data on
absorption near the Galaxy, constraining the temperature and density tightly
(log T (K) = 5.5 − 5.7 and ne = 10−4.7 − 10−3.9 cm−3 when the OviKα measurement
is included), which are conditions suggestive of the local group intergalactic medium
and require supersolar [Ne/O]. Here we first examine the assumptions that have led
us to these results (§2.4.1), and then we discuss their implications for the location of
the absorbing gas (§2.4.2), subject to these caveats.
23
Page 39
2.4.1. Potential Caveats
The Ovi Discrepancy
The interpretation of the UV and X-ray data are particularly important, since
(as Figure 2.5 shows) the combined Oviii/Ovii and Ovi/Ovii ratios can provide
tight constraints on the absorber temperature and density simultaneously (see also
Figure 5 in Mathur et al. 2003). However, in this case the Ovi column density
inferred from the OviKα is a factor of ∼ 4 higher than the combined 1032 A low– and
high–velocity components. Since both the X-ray and UV transitions trace the same
atomic state, the inferred column densities should match. A similar disagreement
has been seen in intrinsic AGN absorption systems (see Krongold et al. 2003;
Arav et al. 2003); however, in these cases it is typically attributed to saturation
or a velocity–dependent covering factor, neither of which is relevant to this z ≈ 0
absorption.
On the other hand, our OviKα measurement provides a test for these
attributions; the local absorption, after all, is likely a dramatically different
physical system than an AGN outflow, yet the same conflict arises. A macroscopic
explanation does not adequately describe how this discrepancy is seen in both
physical systems, so the actual reason may lie in the atomic physics of highly ionized
plasmas. For example, some fraction of the Ovi may be excited through collisions or
recombination from Ovii, and thus unable to produce 1032 A absorption while still
24
Page 40
absorbing OviKα photons. While a scenario that produces significant depopulation
of the Ovi ground state is difficult to envisage in such a low density plasma, we
are investigating further the statistical equilibrium of Ovi including photoexcitation
and recombination in order to study such effects in more detail. However, it
should be emphasized that this is not an isolated case so there must be a physical
explanation for the Ovi discrepancy, and the resolution of this paradox is crucial to
our understanding of Ovi UV and X-ray absorption and how it relates to the Ovii.
There is also the possibility that the line was misidentified as OviKα, and
is actually another intervening Ovii absorption line at z = 0.0195. This latter
explanation is unlikely since no other absorption lines at this redshift are seen in the
FUSE or Chandra spectra; additionally, this would require the line to fall exactly
on the Ovi rest wavelength, which seems like an improbable coincidence. Another
possibility is that the theoretical oscillator strength of the OviKα transition is
incorrect, but the value given in Pradhan (2000) would need to be low by a factor
of ∼ 2 − 4, in sharp contrast to the successful calculations of flu for inner shell
transitions in other ions in the same paper. Nevertheless, due to the discrepancy
between the UV and X-ray Ovi column density measurements, we present both
possibilities: either (a) the OviKα line measures NOVI, or (b) it does not and is thus
ignored.
25
Page 41
Absorption Components
The Doppler parameter measurements indicate the existence of two distinct
components along the Mkn 421 line of sight: one seen in the thick–disk OviLV
1032 A absorption with bLV = 80.6 ± 4.2km s−1, and the Ovii absorber with
bOVII = 31 − 46km s−1 (1σ limits). The Ovi HVC may represent a third phase (if
case (b) above is correct) with bHVC = 35+18−10km s−1 (from the FWHM measurement).
This agrees surprisingly well with the Ovii b measurement, and is consistent with
numerical simulations of the nearby IGM (Kravtsov et al. 2002). However, the
extremely low OviHVC/Ovii ratio requires a temperature much higher than the
upper limit provided by bOVII. In order for the HVC to trace the same gas as
Ovii (case a), then, the aforementioned atomic physics effects would need to be
suppressing Ovi HVC absorption and not the OviLV line. Sembach et al. (2003) list
mean Doppler parameters for a variety of HVCs, both Galactic and probable Local
Group; unfortunately, the dispersion in these values and the errors on bOVII and bHVC
measured here are both too large to associate the components presented here to one
of their classifications.
It is also important to note that our analysis assumes a single phase origin
for the included X-ray lines. This assumption is consistent with the data, given
the good agreement between the three Ovii lines in the calculated ranges of b and
NOVI (Figure 2.3). Even so, if any of the ionic species arises in more than one
26
Page 42
phase along the line of sight, our results could be affected. For example, a Galactic,
purely collisionally ionized medium can in principle reproduce the observed relative
abundances of Oviii, Ovii, OviKα, and Ne ix if several unresolved components
are invoked to explain this absorption. However, the simplest explanation (a
single–phase, low–density, partially photoionized extragalactic absorber) is fully
consistent with all of these line measurements, and the similarity of the derived
absorber properties to expectations for the local WHIM lend further support to this
model (see §2.4.2).
Abundances
Due to the uncertainty associated with the OviKα absorption, metal abundances
relative to oxygen play a particularly important role in this analysis. By adjusting
the metal abundances of the absorbing gas, the consistency of the solutions can in
principle be improved with and without the OviKα measurement. Although the
log ne− log T contour plots are useful for finding solutions, they cannot easily be used
to determine the effects of changes in elemental abundances; thus, Figure 2.9 shows
Ni/NOVII as a function of log T for log ne = −3.9 and Figure 2.10 for log ne = 0. In
these figures, the y–ranges given by the measured ratios (thick lines) shift up and
down as a result of decreases and increases in the abundances relative to oxygen,
respectively; thus different parts of the theoretical curves would be in bold, moving
the allowed temperature ranges (shown in the lower panel) to the left or right. Solar
27
Page 43
abundances here are taken to be the Cloudy 90 defaults (Ferland 1996; Grevasse &
Anders 1989).
In case (a), i.e. if the OviKα measurement of NOVI is correct, then the
temperature and density of the absorber are tightly constrained in a metallicity–
independent manner, and relative abundances for other elements can be estimated.
Under this assumption, the oxygen ion ratios are consistent within a range
of ne = 10−4.7 − 10−3.9 cm−3; however at solar abundances the Ne ix/Ovii
ratio demands higher temperatures than allowed by the OviKα/Ovii ratio.
Over this range of densities, the permitted abundances of neon, carbon, and
nitrogen abundances relative to oxygen (that is, the range of abundances which
produce line ratios consistent with both OviKα/Ovii and Oviii/Ovii) are then
0.6 ≤[Ne/O]≤ 2.2, −0.8 ≤ [C/O]≤ 0.3, and [N/O]≤ 0.9. Note that supersolar
[Ne/O] has also been observed in the z = 0 absorber toward PKS2155–304 (Nicastro
et al. 2002). The improvement in the fit from supersolar [Ne/O] is shown in Fig 2.7.
Note, however, that in this case the discrepancy between the OviKα and 1032 A
measurements becomes even more severe. Since the bulk of the OviLV cannot be
associated with the Ovii due to the different Doppler parameters, the UV Ovi
component associated with the WHIM (hence the OviKα) must be substantially
weaker than the Galactic Ovi absorption; thus the discrepancy is correspondingly
larger.
28
Page 44
On the other hand, if (b) the OviKα line is ignored then the relative abundances
in the absorber can be adjusted such that the measured line ratios are consistent
with either a low– or high–density absorber. As shown in Figure 2.6, a high–density,
collisionally–ionized medium fits the data with solar abundances. Assuming this is
the case, the temperature is then completely constrained by the Oviii/Ovii ratio
at log T = 106.1−6.2 K, and the relative abundances consistent with the Oviii/Ovii
ratio in this temperature range are −0.6 ≤[Ne/O]≤ 0.6, −0.6 ≤[C/O]≤ 0.3, and
[N/O]≤ 0.4.
The requirement that [Ne/O] be supersolar does not affect the viability of the
partially–photoionized model: in both the ISM and low–z IGM, [Ne/O] is observed
to be significantly larger than the solar value (e.g. Paerels et al. 2001; Nicastro
et al. 2005a). This may be an intrinsic property of the enriched gas ejected into
the IGM and ISM, or instead could be due to depletion of C, N, and O onto dust
grains in supernova ejecta or quasar winds (Whittet 1992; Elvis et al. 2002). If the
dust destruction timescale is long and the IGM is continuously enriched by this
latter mechanism, then the observed supersolar [Ne/O] would be expected. All of
these enrichment scenarios are quite uncertain, but few (if any) are able to produce
[Ne/O]< 0. Indeed, the solar neon abundance itself is quite uncertain since it is
inferred from solar wind measurements. The increase in the solar neon abundance
proposed by Drake & Testa (2005) is supported by these observations, and may
provide another physical argument for the lack of subsolar [Ne/O].
29
Page 45
2.4.2. Where does the X-ray absorption originate?
Assuming the Ovii absorption system is homogeneous, its radial extent can be
estimated by calculating r ≈ NH/(µne), where µ ≈ 0.8, log(nO/nH) = −3.13 (solar
abundance), and
NH = NOVII ×(
NO,tot
NOVII
)
× 103.13−[O/H]. (2.2)
The second term in the equation is approximately unity, since over the range
of temperatures and densities implied from the Oviii/Ovii ratio, Ovii
is the dominant ionization state by at least an order of magnitude; thus,
log NH = log NOVII + 3.13 − [O/H] = {20.37 − ([O/H] + 1)} ± 0.11. The
measurement error on NOVII is small compared to the uncertainty range in ne,
so it can be disregarded. Assuming that (a) the OviKα line does measure NOVI,
the 2σ range of densities is −4.7 ≤ log ne ≤ −3.9, resulting in a radial extent of
r = (0.8 − 4.9) × 10−([O/H]+1) Mpc. These radial extents are consistent with those
expected from a Local Group medium or local filament interpretation for this
absorption (Nicastro et al. 2002, 2003) and too large to be confined within a Galactic
halo. The absorber extent could be marginally consistent with a Galactic corona if
the metallicity is high (r = 80 − 490 kpc, 2σ limits at solar metallicity); however,
such a scenario seems implausible, particularly if this corona primarily consists of
gas accreted from the metal–poor IGM.
30
Page 46
It is unlikely that this absorption system, if extragalactic, is spherically
symmetric (particularly in a “local filament” interpretation). The total
mass in the Ovii system can be written as Mtot = f × (4/3)πr3(1.4nHmH),
where f (< 1) parameterizes the departure from spherical symmetry and
1.4nHmH ≈ nHmH + nHemHe. Replacing r with NH/(µne) and plugging in relevant
values,
Mtot = 9.9 × 1012M⊙(
10−3([O/H]+1))
(
ne
10−4 cm−3
)
−2
f (2.3)
This is several times larger than estimates of the total Local Group binding mass,
e.g. Mtot = (2.3± 0.6)× 1012M⊙ as calculated by Courteau & van den Bergh (1999).
Baryonic matter should only contribute ∼ 15% of this mass (if the baryon–to–dark
matter ratio is equal to the cosmological value), so our estimate appears high. This
discrepancy can be easily resolved with different values of [O/H] and f . For instance,
if we assume an oxygen abundance of 0.3 times solar rather than 0.1, then the range
of possible masses becomes 2.0 × 1011fM⊙ ≤ Mtot ≤ 7.9 × 1012fM⊙. Thus, unless
[O/H] is very high or f ≪ 1, the Ovii absorber almost certainly accounts for a
major fraction of the baryonic matter in the Local Group. This sightline may also
be probing gas that is not gravitationally bound to the Local Group (i.e., the Local
Filament), which may explain why our range of Mtot extends to such high values.
Although Collins et al. (2005) argue that a Local Group origin requires that the
31
Page 47
Ovii absorbers contain too much mass, we see here that the total mass is in fact
consistent with expectations.
If this absorption does only trace the Local Group medium, then
constraints on the extent of the absorber can be derived by assuming
Mtot = 0.15MLG ≈ 3.5×1011M⊙. In this case, r3f = 3Mtot/(4π×1.4µnemH). Taking
the 2σ upper limit of log nH = −3.9, the 2σ lower limit on MLG = 1.1 × 1012M⊙,
and assuming f = 1, we obtain a lower limit of r > 0.2 Mpc for the Ovii absorption.
Similarly, the upper limit on the radius is r < 0.6f−1/3 Mpc. This seems somewhat
small compared to the actual size of the Local Group, but once again is dependent
on the geometry of the absorber. A value of f ≈ 0.1 brings this upper limit more in
line with the Mpc scales expected in the Local Group; this may indicate that the
Mkn 421 line of sight probes an extended, filamentary WHIM distribution. This
calculation also assumes that the density of the Local Group medium is constant
with radius, when the actual density profile is more likely centrally concentrated.
These measurements are also affected by the Ovi discrepancy: in case (b), only the
lower density limit of log(ne) > −4.7 (from the 2σ Ovi Kα upper limit) applies, so
the upper radius and mass limits are still valid. Nevertheless, the consistency with
the expected Local Group parameters is intriguing.
32
Page 48
2.4.3. Comparisons to Other Studies
Kravtsov et al. (2002) used constrained simulations to study the properties
of the Local Supercluster region; sky maps produced from this simulation (their
Figure 5) show filamentary structures near the Mkn 421 direction, possibly
corresponding to the observed absorption. Additionally, they note that a Local
Group medium would exhibit a low Doppler parameter (b ∼< 60 km s−1) out to
distances of ∼ 7 Mpc, consistent with our Ovii measurement.
The inferred properties of the X-ray absorption along this line of sight also
appear similar to other observations: toward 3C 273, for example, Fang et al. (2003)
find 5.36 < log T < 6.08 and comparable Ovii column density; however, their
inferred Ovii Doppler parameter is significantly higher than that toward Mkn 421:
b > 100km s−1. On the other hand, the z = 0 absorber toward PKS 2155 − 304
(Nicastro et al. 2002) exhibits a temperature consistent with the Mkn 421 absorber,
yet inferred density about an order of magnitude lower. Compared to the two
intervening filaments seen toward Mkn 421 (Nicastro et al. 2005a), the density of
the z = 0 absorption agrees with the derived lower limits (log ne ∼> −5 for both
filaments), but the filaments appear to exhibit higher temperatures than that
derived for the local absorption. These variations along different lines of sight
simply demonstrate the complex nature of the absorption, and the diversity of
33
Page 49
temperature and density environments produced in the structure formation process
(e.g., Kravtsov et al. 2002).
The detection of X-ray absorption lines toward the Large Magellanic Cloud
binary LMC X-3 by Wang et al. (2005) presents an important consideration for these
results as well. The measured Ovii and Ne ix equivalent widths and upper limits on
Oviii and OviiKβ are all consistent with the same lines measured toward Mkn 421
(albeit with much larger statistical errors). Although this detection provides evidence
of a hot intervening absorber between the Galaxy and LMC, it certainly does not
rule out a primarily extragalactic origin for the Mkn 421 absorber. Any absorption,
either Galactic or extragalactic, is likely to be inhomogeneous; thus, it’s entirely
plausible that the LMC X-3 sightline probes hot Galactic gas (perhaps enhanced by
winds or outflows from both the Galaxy and LMC), while the absorption toward
Mkn 421 is primarily due to low–density nearby WHIM gas.
2.5. Summary and Future Work
Through long–duration Chandra and FUSE observations of Mkn 421 in
outburst, we have obtained unprecedented measurements of a variety of z ≈ 0
absorption lines, many of which likely arise in extragalactic, partially photoionized
gas. A brief summary of our results follows.
34
Page 50
1. The relative strengths of the three Ovii K-series lines imply 2σ Doppler
parameter constraints of 24 < b < 55 km s−1. This is inconsistent with the
value of b = 80.6 ± 4.2 km s−1 derived for the Galactic low–velocity Ovi,
indicating that the OviLV and Ovii likely arise in different phases. The Ovii
b value is, however, consistent with the local IGM simulations of Kravtsov et
al. (2002) out to distances of several Mpc.
2. A weak high–velocity Ovi 1032 A component also appears in the FUSE
spectrum. Although its width is consistent with the Ovii b measurement, the
OviHVC column density is too low to be associated with the Ovii absorption
unless T ∼> 107 K (which itself is ruled out by the upper limit on b). The
OviHVC may thus represent a distinct third component along this line of sight.
3. The column density inferred from the OviKα line is a factor of ∼ 4 higher
than that measured from the Ovi 1032 A transition. This may be due to
unaccounted–for atomic physics effects, in which case the Kα line may provide
a more accurate measurement of NOVI than the 1032 A line. We consider both
cases:
(a) If the Ovi Kα line measures NOVI, then strong constraints on the
temperature, density, and relative abundances of the X-ray absorber can
be derived: T = 105.5−5.7 K and ne = 10−4.7 − 10−3.9 cm−3, with allowed
abundances of 0.6 ≤[Ne/O]≤ 2.2, −0.8 ≤[C/O]≤ 0.3, and [N/O]≤ 0.9
35
Page 51
(all 2σ ranges). This range of densities, combined with NOVII, implies a
total mass and extent consistent with those expected in the Local Group
and/or Local Filament if the gas metallicity is low. However, in this case
the Ovi UV–X-ray discrepancy becomes worse since (due to the Doppler
parameter constraints) only a small portion of the OviLV line can be
associated with the extragalactic X-ray lines.
(b) If, instead, the Ovi Kα line does not correctly measure NOVI, then the
Ovi associated with the Ovii absorption must be fully blended with
the Galactic 1032 A Ovi, and thus not measurable. In this case a lower
density limit of log ne > −4.7 is found, which is consistent with either a
Galactic or extragalactic medium.
Much work remains to be done—both in order to better understand the data
presented here, and to determine the true nature of the z ≈ 0 X-ray absorption.
Higher signal to noise data along the Mkn 421 sightline would be useful to obtain
better column density constraints, particularly on Oviii, OviKα, and Ne ix, and
thus better constrain the effects of photoionization on the absorber. Data of
comparable quality along other sightlines would be invaluable as well, both to probe
other regions surrounding the Galaxy and to reconfirm the tantalizing Ovi results
presented herein. Higher–resolution simulations of the Local Group may allow us to
determine the ionic column densities expected in the WHIM, and thus whether or
not the observed absorption can possibly arise in the WHIM. Finally, more detailed
36
Page 52
modeling of Ovi inner–shell transitions would shed a great deal of light on whether
or not the Ovi X-ray/UV discrepancy is real, and thus provide an invaluable
framework for studying new (and existing) X-ray data.
37
Page 53
Line ID λresta λobs
b ∆vFWHM vobs Wλc log Ni
c,d Note
(A) (A) (km s−1) (km s−1) (mA)
X-ray:
Ar XVKα 24.737 24.737 · · · · · · < 3.09 < 15.12
C VKα 40.268 40.26 ± 0.02 · · · −60 ± 150 11.3+3.3−2.6 15.19 ± 0.15
C VIKα 33.736 33.736 ± 0.02 · · · 0 ± 180 7.2 ± 1.4 15.31 ± 0.11
Ne IXKα 13.447 13.431 ± 0.02 · · · −360 ± 450 2.4+0.9−0.8 15.48 ± 0.24
Ne XKα 12.134 12.11+0.03−0.02 · · · −590+740
−490 < 5.04 < 16.21 1
N VIKα 28.787 28.755 ± 0.02 · · · −330 ± 210 4.1 ± 1.5 15.02+0.19−0.24
N VIIKα 24.781 24.781 · · · · · · < 2.86 < 15.16
O VKα 22.374 22.374 · · · · · · < 2.20 < 14.97 2,3
O VIKα 22.019 22.023 ± 0.02 · · · 50 ± 270 2.4 ± 0.9 15.07+0.17−0.22 2
O VIIKα 21.602 21.603 ± 0.02 · · · 10 ± 280 9.4 ± 1.1 16.22 ± 0.23
O VIIKβ 18.629 18.612 ± 0.02 · · · −273 ± 320 4.6 ± 0.7 16.28 ± 0.13
O VIIKγ 17.768 17.762 ± 0.02 · · · −100 ± 340 2.0 ± 0.7 16.19+0.16−0.21
O VIIIKα 18.969 18.974 ± 0.02 · · · 80 ± 320 1.8 ± 0.7 15.17+0.16−0.24
(cont’d)
Table 2.1. Observed z ≈ 0 lines.
38
Page 54
Table 2.1—Continued
Line ID λresta λobs
b ∆vFWHM vobs Wλc log Ni
c,d Note
(A) (A) (km s−1) (km s−1) (mA)
UV:
O VI1032 1031.926 1031.88 ± 0.01 152.3 ± 7.0 −13.7 ± 2.6 270.9 ± 7.9 14.43 ± 0.02
O VI1032, HVC 1031.926 1032.30 ± 0.04 59+31−17 108 ± 12 18.5 ± 5.6 13.18+0.12
−0.16
O VI1037 1037.617 1037.59 ± 0.01 145.4 ± 11.6 −9.3 ± 4.0 164.9 ± 8.4 14.47 ± 0.03
aFrom Verner et al. (1996), except Ovi and Ov which are from Schmidt et al. (2004)
bWavelength uncertainty is measured from fit or intrinsic LETG 0.02 A error, whichever is greater. For
upper limits, the line position was frozen to the rest wavelength.
cError bars are 1σ; upper limits are 2σ.
dColumn densities are calculated using curve–of–growth analysis with b = 80.6± 4.2km s−1 for Cv, Nvi,
and Ov, and the Ovii Doppler parameter of b = 31 − 46km s−1 (1σ range) for all other lines.
Note. — (1) The Ne X line lies within a detector feature, so only an upper limit on its equivalent width
is given. (2) The Ovi and Ov λrest values are from laboratory measurements by Schmidt et al. (2004);
theoretical oscillator strengths are taken from Pradhan et al. (2003). (3) There is an absorption line at
−760km s−1 from the Ov rest wavelength, but this is more likely Ovii associated with Mkn 421.
39
Page 55
Fig. 2.1.— Portions of the Mkn 421 Chandra LETG spectrum (points) and the
best–fitting model (histogram). Absorption lines at z ≈ 0 are labeled, and vertical
tick marks indicate absorption from the z = 0.011 (solid) and z = 0.027 (dotted)
intervening WHIM filaments (Nicastro et al. 2005a).
40
Page 56
Fig. 2.2.— FUSE spectrum of Mkn 421 around the O VIλλ1032, 1037 A absorption
doublet (histogram). The dark curve shows the best–fit model with (solid line) and
without (dotted line) the 1032 A HVC included. The inset plot shows the best–fitting
Galactic (large Gaussian) and high–velocity (small Gaussian) components for the
1032 A line, plotted against velocity relative to the Ovi rest wavelength.
41
Page 57
Fig. 2.3.— Contours of allowed NOVII and b for the Ovii Kα (yellow), Kβ (red),
and Kγ lines. Shaded regions depict the 1σ errors on Wλ, with the best-fit Wλ line
in the center of each region. The overlap between these three regions 2σ limits of
24km s−1 < b < 55km s−1; the blue box depicts the 1σ ranges in log(NOVII) and b.
Also labeled on the top axis is log Tmax, the maximum temperature for a given b value.
42
Page 58
Fig. 2.4.— Contours of allowed NOVI and b for the Ovi 1032 A (green), 1037 A
(red), and putative Kα 22 A (yellow) lines. Contour (a) is derived from the 1032 A
OviLV equivalent width while (b) is from the measured FWHM; the intersection
between the two green contours provides a tight constraint of b = 80.6 ± 4.2km s−1
and log NOVI(cm−2) = 14.432 ± 0.016 for the OviLV (shown as the blue cross). Note
that the Ovi Kα transition predicts NOVI about 0.5 dex higher than the UV line,
and this discrepancy cannot be explained by saturation.
43
Page 59
Fig. 2.5.— Contours of constant abundance ratios for X-ray and UV oxygen
absorption lines. Vertical bars denote the 2σ range of temperatures inferred from
the abundance ratio at each step in log ne; the black cross shows the “overlap” region
between these contours. The horizontal dashed line is the 2σ upper limit on the
temperature of the Ovii absorber from the Doppler parameter measurement.
44
Page 60
Fig. 2.6.— Same as Figure 2.5, but for ratios of several different ion abundances
to Ovii: OviKα (green), Oviii (red), Cvi (blue), Ne ix (cyan), and Nvii (dotted
black region). The black cross shows the range of log T and log ne derived from
the Oviii/Ovii and OviKα/Ovii ratios. Assuming solar abundances, the OviKα
and Ne ix contours are inconsistent for log ne ∼> −5, while all other ratios (except
OviKαOvii) are consistent at log ne ∼> −4.5. High–density models agree with the
data only if the OviKα measurement is disregarded.
45
Page 61
Fig. 2.7.— Same as Figure 2.6, showing how a neon abundance shift of [Ne/O]= 1
produces better agreement in the low–density regime.
46
Page 62
Fig. 2.8.— Models of ionic column density ratios for ions likely to arise in the Galactic
ISM (assuming pure collisional ionization). Calculated ratios are shown as thin lines
with ±2σ measurements overplotted (thick segments). Upper limits are shown as
dashed lines, and the temperature ranges derived from the different ion ratios are
shown in the lower panel.
47
Page 63
Fig. 2.9.— Same as Figure 2.8, for the ions likely originating in an extragalactic
medium, with log ne = −3.9. Solar abundances relative to oxygen are assumed, and
all calculated ratios are relative to NOVII. The low–density case produces a consistent
temperature solution of T ≈ 105.7 K (from the overlap between the Oviii and OviKα
bold regions), and [Ne/O] must be supersolar.
48
Page 64
Fig. 2.10.— Same as Figure 2.9 but assuming high density with only collisional
ionization. This produces a consistent solution with solar abundances but the
temperatures implied by the Oviii/Ovii and OviKα/Ovii ratios are inconsistent.
49
Page 65
Chapter 3
The Markarian 279 Sightline
While the origin of z ≈ 0 X-ray absorption remains ambiguous, the locations
of lower–ionization components in or near the galaxy, such as the ubiquitous neutral
hydrogen high–velocity clouds (H I HVCs) and associated UV absorption lines (such
as Ovi), are also as yet undetermined. Mrk 279, a nearby (z = 0.03), X-ray bright
Seyfert galaxy, lies in the direction of the H I HVC Complex C, thus providing a
particularly valuable background source that can be used to study these gaseous
components in X-ray absorption. In this chapter I discuss my analysis of deep
Chandra and FUSE spectra of this object, the detected UV and X-ray absorption,
and the implications for gas in the Galaxy and Local Group.
3.1. Data Reduction and Measurements
3.1.1. Chandra
Seven observations of Mrk 279 taken with the Chandra High Resolution Camera
Spectroscopic array (HRC-S) and Low Energy Transmission Grating (LETG), all
50
Page 66
taken in May 2003 and totaling 340 ks of exposure time, were available in the
Chandra archive. Each of these data sets was processed, and instrument response
files built, using the standard data reduction threads for HRC-S/LETG1 with
version 3.3 of the Chandra Interactive Analysis of Observations (CIAO) software
and Chandra Calibration Database version 3.2.0. Since the HRC-S does not have
sufficient energy resolution to distinguish LETG spectral orders, higher orders
can increase the apparent flux at long wavelengths2. This effect was mitigated by
including response files for orders −6 through +6 in our analysis; the inclusion
of orders beyond these had an insignificant effect on the computed instrumental
response.
The seven observations were coadded for a final (unbinned, with ∆λ = 0.0125 A)
signal–to–noise ratio of S/N ≈ 6.5 near 22A. We used the spectral fitting program
Sherpa to fit a power law and foreground Galactic absorption to the spectrum
over 10 − 100 A band (excluding the 49 − 57.5 A and 60.5 − 67.5 A chip gap
regions). The relative Galactic abundances of carbon, nitrogen, oxygen, and neon
were left as free parameters in order to produce a better fit around the absorption
edges. A power–law slope of Γ = 2.3 and equivalent hydrogen column density of
(1.78± 0.03)× 1020 cm−2 is derived, agreeing reasonably well with the Elvis, Wilkes,
& Lockman (1989) value of 1.64 × 1020 cm−2 near this sightline. The best–fit
abundances of N and Ne were approximately zero, oxygen equal to the solar value,
1Available at cxc.harvard.edu/threads/gspec.html
2See cxc.harvard.edu/ciao/threads/hrcsletg orders/
51
Page 67
and carbon 0.15 solar, though we re–emphasize that these do not reflect actual
Galactic abundances but rather provide a better fit near absorption edges where
the calibration is uncertain. A few weak, broad residuals remained afterward,
probably due to calibration uncertainties or source variability; these were corrected
by including four broad Gaussians in the source model (analogous to the technique
described in Nicastro et al. 2005a).
Once the continuum was established, we visually searched the spectrum in
∼ 3 A windows for narrow (unresolved) absorption lines, fitting each one with a
Gaussian. Although several strong lines such as Cvi, Ovii, and Nvii are apparent
at the AGN redshift (z = 0.03), at z = 0 (v ∼< 700 km s−1) only Ovii Kα λ21.602 A
is unambiguously detected at 21.619 ± 0.009A (v = 236 ± 125 km s−1) with an
equivalent width of 26.6 ± 6.2 mA. Upper limits were measured for the Ovii Kβ
line as well as several other ionic species of interest; these measurements are listed
in Table 3.1. Although they are included in the model to improve consistency, the
absorption intrinsic to Mrk 279 and the Galactic interstellar O I lines are not the
focus of this work and will not be discussed further. The resulting fit and residuals
are shown in Figure 3.1 (and a zoomed–in figure near the Ovii wavelength in
Figure 3.2); with a reduced χ2 value of 0.89, the model appears to fit the data quite
well.
The Chandra HRC-S/LETG wavelength scale is uncertain for several reasons,
primarily because of non–linearities resulting from bad amplifiers on the HRC-S
52
Page 68
detector3. While the newly–released CIAO 3.3 software includes a routine to correct
these non–linearities and has reduced the dispersion in wavelength errors to ∼ 6 mA
at short wavelengths in calibration spectra, this routine is in the early stages of
development and wavelengths of individual emission and absorption lines may still
be systematically skewed. However, any systematic wavelength errors should not
vary between observations as long as the telescope pointing is nominal. Furthermore,
while serious wavelength errors are known to occur around 18A, no bad amplifiers
are expected to significantly affect the dispersion relation near 21.6A (J. J. Drake
and N. Brickhouse, private communication).
To check the absolute wavelength scale near the Ovii line, we retrieved the
nearest HRC–S/LETG calibration observation of the X-ray bright star Capella
(observation 3675, taken on 2003 September 28) from the Chandra archive and
reprocessed the data in exactly the same manner as the Mrk 279 data. The
wavelength of the strong Ovii emission line was found to be 21.606 ± 0.002A or
56 ± 28 km s−1, which is consistent with the +30 km s−1 radial velocity of Capella
as listed in the SIMBAD database4. As a separate check, we reduced the Mrk 279
Chandra data both with and without the wavelength correction routine; the
difference in the measured Ovii wavelength between the two was only 4 mA, much
lower than the statistical error on the line position. Thus, although the possibility of
3See cxc.harvard.edu/cal/Letg/Corrlam/
4simbad.u-strasbg.fr/sim-fid.pl
53
Page 69
systematic wavelength errors must be kept in mind, it appears as though such effects
are insignificant compared to the statistical error on the measured Ovii wavelength.
3.1.2. FUSE
Mrk 279 was observed four times with FUSE between December 1999 and
May 2003 with a total exposure time of 224 ks; all calibrated data from these
observations were obtained through the Multimission Archive at STScI website.5
To account for small shifts in the FUSE wavelength scale during the observations,
each of the constituent exposures was cross–correlated over the 1030 − 1040A range
and the relative positions of strong absorption lines were checked by eye. The data
from 18 May 2002 were not of sufficient quality to reliably perform this wavelength
calibration and were thus excluded. The coadded, wavelength-shifted spectra from
each observation were then cross–correlated with each other, scaled so that their
continuum intensities matched, combined and rebinned by five pixels (∼ 10 km s−1)
to produce a final spectrum with S/N ≈ 27 near 1032A and an effective exposure
time of 177 ks.
To account for possible systematic offsets, the absolute wavelength scale of
the final spectrum was corrected following the method employed by Williams et
al. (2005): the Galactic Si II λ1020.699 and Ar I λ1048.220 absorption profiles
5archive.stsci.edu/
54
Page 70
were fit with multiple Gaussian components and the column density–weighted
average velocities calculated. These were found to be −70.0 km s−1 and −65.5 km s−1
respectively, while the average velocity of the Wakker et al. (2003) multi–component
fit to the Galactic H I toward the Mrk 279 sightline is v ≈ −37 km s−1. Since these
low–ionization lines are expected to co-exist with the H I, a +30 km s−1 shift was
applied to the wavelength scale of the FUSE spectrum.
The final combined and calibrated FUSE spectrum shows strong Ovi absorption
from the Galactic thick disk at v ≈ 0 and a weaker Ovi high velocity component
(OviHV) at v ≈ −150 km s−1(Figure 3.3). Each of these features was initially
modeled with a single Gaussian; however, this provided a poor fit for the thick–disk
absorption due to an asymmetric blue wing on the line profile; thus, the Galactic
Ovi was instead fit with a broad and a narrow Gaussian component (hereafter
denoted OviB and OviN respectively). Measured line parameters are also listed
in Table 3.1; note that the velocity of the OviHV is tightly constrained by this
fit as v = −159.9 ± 2.6 km s−1, which is inconsistent at the 2.8σ level with the
Ovii velocity of 236 ± 125 km s−1 (assuming the HRC-S/LETG wavelength scale is
correct), indicating that these two components almost certainly are not related.
The total equivalent widths of the Ovi Galactic and high–velocity components
are 262.5±6.7 mA and 43.6±3.5 mA respectively, in agreement within the estimated
systematic errors of the Wakker et al. (2003) measurements of 247 ± 8 ± 25 mA and
53±6±17 mA (where the first and second error values are statistical and systematic,
55
Page 71
respectively). As was the case for the Mrk 421 sightline (Williams et al. 2005),
our fitted Galactic+HVC Ovi equivalent width agrees quite well with the Wakker
et al. (2003) direct–integration value but our measurement of the OviHV strength
is somewhat lower and Galactic absorption higher; this is most likely because our
method better accounts for blending between the Galactic and high–velocity Ovi in
cases where distinct absorption components are evident (as discussed in more detail
in Williams et al. 2005).
3.2. Analysis
3.2.1. Doppler Parameters and Column Densities
In order to derive physical properties of the observed absorption, it is necessary
to first convert the measured equivalent widths into ionic column densities Ni. This
requires knowledge of the Doppler parameter b, since at a fixed column density Wλ
decreases for lower values of b. However, the nominal LETG resolution of 0.05A
(∼ 700 km s−1 at 21A) precludes direct measurement of the Ovii line width. A
method similar to that employed in Williams et al. (2005) is thus used to place
limits on the Ovii Doppler parameter using the measured equivalent width of the
Ovii Kα line and the upper limit on Ovii Kβ. Equivalent widths and FWHM
values for each transition were calculated (assuming a Voigt absorption profile) over
56
Page 72
a grid of NOVII and b, and the tracks in the NOVII − b plane consistent with each
measured 2σ equivalent width limit are plotted in Figure 3.4.
Determining the ranges of NOVII and b for which the measured column densities
are consistent with each other (i.e. the area over which the Kα and Kβ tracks
“overlap”) is more complicated. In the Mrk 421 data the first three lines in the Ovii
K–series were individually detected at ≥ 3σ confidence. In that case, the differences
between pairs of predicted NOVII values (and the joint errors on the differences) were
calculable, allowing easy determination of the region over which the three transitions
predicted consistent NOVII and b values. In the case of Mrk 279, however, this same
method cannot be used because while the Ovii Kα line has been detected, only an
upper limit is available for the Kβ line.
However, since the absorption line properties for various column densities and
Doppler parameters are known, limits on these quantities can be determined using
the Chandra spectrum itself. For each point in the NOVII − b plane, Ovii Kα and
Kβ absorption lines with the calculated Wλ and FWHM values were added to the
best–fit continuum model, and the χ2 statistic calculated using the “goodness”
command in Sherpa. Over the calculated parameter ranges, the minimum χ2
point was on the b = 200 km s−1 boundary. Since the Kβ line was completely
undetected (with a best–fit amplitude near zero), the χ2 value asymptotically
approaches a minimum as b increases. We thus assumed a minimum χ2 value from
a fit with a fixed unsaturated line ratio (Wλ(Kβ) = 0.15 ∗ Wλ(Kα)) and calculated
57
Page 73
∆χ2 = χ2(NOVII, b) − χ2min for every point. The 95% confidence interval (∆χ2 < 6)
determined with this method is shown in Figure 3.4; at this confidence level all
Doppler parameters between 24 < b < 74 km s−1 are ruled out.
Although these curve–of–growth diagnostics can in principle also be applied
to the UV Ovi λλ1032, 1038 absorption doublet, it is typically more difficult
because few Galactic Ovi sightlines are sufficiently saturated to significantly affect
the doublet line ratio; additionally, the close proximity of the 1037A line to other
Galactic absorption features makes deblending difficult (Wakker et al. 2003; Williams
et al. 2005). However, since the Ovi λ1032 line is fully resolved in the FUSE
spectrum, the measured line width is highly sensitive to b while the equivalent width
traces NOVI. While Wλ decreases for a saturated line of a fixed column density,
the observed FWHM increases from the unsaturated value of FWHM = 1.665b. To
account for these saturation effects, we computed Ovi λ1032 equivalent widths and
FWHM values over a grid of NOVI and b. Regions for which the FWHM and Wλ
measurements are consistent with predicted values are shown for both OviB and
OviN in Figure 3.5; since the FWHM and Wλ regions overlap nearly orthogonally,
strong constraints are placed on the column density and velocity dispersion of the
Ovi–bearing gas. None of the Ovi broad, narrow, or high–velocity components
are significantly saturated, with derived Doppler parameters of 61.5 ± 3.5 km s−1,
38.8 ± 2.8 km s−1, and 32.0 ± 4.6 km s−1 (1σ errors) respectively.
58
Page 74
All Ovi component Doppler parameters are inconsistent with the upper allowed
range (b > 74 km s−1) determined from the Ovii ratios, and only the OviHV is
marginally consistent with the lower Ovii range (b < 24 km s−1). This indicates that
the majority of the v ≈ 0 Ovi cannot originate in the same phase as the observed
Ovii absorption. Although at first glance it appears possible for the OviHV and
Ovii to coexist, the velocity separation of the Ovii and OviHV (as discussed
in §3.1.2) makes this interpretation unlikely; additionally, the large Ovii/OviHV
column density ratio (NOVII ∼> 17.7; Figure 3.4) requires high temperatures which
are in turn ruled out by the low b value (discussed further in the following section).
It is also possible that the thick–disk Ovi consists of only one component with a
non–Gaussian shape, in which case the actual velocity dispersion could be larger.
Approximating the low–velocity Ovi with a single Gaussian component yields
b ≈ 75 km s−1— barely consistent with the 95% lower limit on the Ovii Doppler
parameter. We thus conclude that the Ovii most likely does not coexist with any of
the measured Ovi components.
3.2.2. Temperature and Density Diagnostics
The derived ionic column density ratios (and upper limits thereupon) can be
used to determine the physical state of the absorbing medium. Since the extent, and
hence the density, of the absorber is unknown, we cannot assume a priori whether or
not photoionization from the Galactic or extragalactic background plays a significant
59
Page 75
role in the ionization balance of the medium. Specifically, in a higher–density, hot
medium (such as a Galactic corona), collisional ionization would be the dominant
process by far, while in a low–density WHIM scenario photoionization is expected to
play a significant role. Including the effects of photoionization along with collisional
ionization is crucial in order to most accurately determine the state of the gas (cf.
Nicastro et al. 2002; Mathur et al. 2003; Williams et al. 2005).
We used version 05.07 of the ionization balance code Cloudy (Ferland et
al. 1998) to calculate ionic abundances for all measured elements over a grid of
T = 104.5−107.4 K and ionization parameters log(U) = −6.3 to 0.7 (where U = nγ/ne
is the ratio of the number densities of ionizing photons and electrons in the plasma),
with a step size of 0.1 dex in both log T and logU . The Sternberg et al. (2002)
fit to the metagalactic radiation field was assumed; this is based on the observed
background from infrared to X-rays (except for the unobservable radiation near
the Lyman limit, which is taken from the theoretical model of Haardt & Madau
(1996)). The normalization of this background corresponds to electron densities of
ne = 10−7 − 1 cm−3 over the calculated log(U) range — i.e., log(ne) = −6.3− log(U).
As equivalent widths of different transitions from the same ion can be used
to place limits on the absorber’s velocity dispersion and column density (§3.2.1),
any measured column density ratio defines a track in the log T − log U plane. The
overlap between two or more such tracks (derived from different ion ratios) can
then be used to place constraints on the gas temperature and density. Although in
60
Page 76
principle the ratio between any two ionic species can be employed, column density
ratios between different ions of the same atomic species are metallicity–independent
and hence produce the strongest diagnostics. Since only Ovii is strongly detected at
z = 0 in the Mrk 279 Chandra spectrum and Ovi, Ovii, and Oviii together provide
strong temperature and density constraints (Mathur et al. 2003), we focus primarily
on these ions.
A limit on NOVIII can be easily determined from the Oviii equivalent width
upper limit and assuming the Ovii Doppler parameter; however, including the Ovi
column density is less straightforward due to the presence of multiple components.
Since the Ovii Doppler parameter is inconsistent with both of the broad v ≈ 0 Ovi
components, there are two likely possibilities: (1) the Ovii Doppler parameter is
actually in the lower range (b < 21 km s−1) and the Ovii is associated with the
OviHV, or (2) the Ovi absorption produced by the Ovii–bearing gas is too weak
to be detected in our FUSE spectrum, so only an upper limit on NOVI can be used
in this analysis. The former case is highly unlikely — not only is the centroid of
the Ovii line inconsistent with the velocity of the OviHV, but such a low Doppler
parameter requires extremely high Ovii column densities (NOVII ≈ 1018 cm−2).
This in turn produces an extremely large NOVII/NOVI ratio which requires high
temperatures (T > 107.4 K, the upper limit of our calculation). Since the OviHV
Doppler parameter implies a maximum temperature of Tmax ∼< 106 K, such an
association appears impossible.
61
Page 77
If, instead, the associated Ovi absorption is too weak to be detected, then this
absorption most likely takes the form of a broad (b ∼> 80 km s−1, from the Ovii b
limit) absorption line superposed on the v ≈ 0 Ovi. Such a line was included in the
FUSE spectrum fit and 2σ upper limits calculated for a velocity dispersions of 100
and 200 km s−1. The temperatures and densities consistent with the Oviii/Ovii
and Ovi/Ovii upper limits for these values of b are shown in Figures 3.6 and 3.7
respectively. The Oviii/Ovii ratio sets an upper limit of log T ∼< 6.3 and a minimum
density of 10−6 cm−3 in both cases. For b = 100 km s−1 the limits set by NOVI/NOVII
are inconsistent with the NOVIII/NOVII ratio, but for b = 200 km s−1 the contours
begin to overlap with 5.9 < log T < 6.3 and log ne > −5.1. Thus, if the Ovii is
associated with a broad undetected Ovi component, a large Doppler parameter is
required to reconcile the oxygen ion ratios. This, along with the inferred temperature
and density limits, are both consistent with expectations for the local WHIM, though
the large velocity dispersion compared to the upper temperature limit derived from
the Oviii/Ovii ratio indicates that the line broadening is primarily nonthermal.
Although a similar analysis can be performed with the Nvii/Ovii and
Ne ix/Ovii ratios, the derived limits are in all cases weaker than those set by the
Oviii/Ovii upper limit and have been excluded from Figures 3.6 and 3.7 for the
sake of clarity.
62
Page 78
3.2.3. The AGN Warm Absorber
The Chandra spectrum of Mrk 279 shows strong Ovii, Oviii, Cvi, and Nvii
absorption at a redshift consistent with the AGN (z = 0.03). It is thus possible
that the Ovii absorption line at z = 0 is contaminated, either by Ovii outflowing
from Mrk 279 or from another absorption line at z = 0.03. The former scenario is
probably not the case since this would require both an unlikely coincidence of the
outflow velocity with the AGN redshift (v ≈ 9000 km s−1). To check the latter case,
we used the PHASE model (described by Krongold et al. 2003) to fit the intrinsic
absorption in a self–consistent manner. With this fit we found that the redshifted
Nvii Kβ line falls at 21.5A, near the Ovii Kα rest wavelength but well outside the
line profile, and is weak enough that its effect on the Ovii Kα equivalent width
measurement is most likely negligible. No other warm–absorber lines are expected
near 21.6A for outflow velocities between zero and 9000 km s−1; thus, the z = 0 Ovii
measurement is unlikely to be contaminated by any lines from the warm absorber.
The details of the warm absorber model are the subject of a forthcoming paper
(D. L. Fields et al., in preparation) and will not be discussed further here.
63
Page 79
3.3. Discussion
3.3.1. Comparison to the Mrk 421 Sightline
Although the quality of the Mrk 279 Chandra spectrum is far lower than that
of Mrk 421 (Williams et al. 2005), the differences between the local absorption
seen along the two lines of sight are striking. While the velocity of the Mrk 421
Ovii absorption is near zero and thus cannot be distinguished from the low– and
high–velocity Ovi seen in the spectrum, the Mrk 279 Ovii and OviHV velocities
are significantly different. Furthermore, the derived Doppler parameters of the
Ovii absorption along these sightlines — 24 < b < 55 km s−1 and b > 74 km s−1 for
Mrk 421 and Mrk 279 respectively — differ substantially. In both cases association
of the Ovii with any Ovi component is ruled out and the derived temperature,
density, and column density limits are consistent with each other (though with large
errors). However, the strong discrepancy between the velocity dispersions of the two
absorbers suggests that their origins may differ. Such a difference in origin may not
be surprising, given the large (∆l ≈ 60◦) separation between the two sightlines.
3.3.2. Origin of the Absorption
The unique properties of the absorption components along the Mrk 279 sightline
provide some tantalizing clues as to the origin of the local host gas. Taking all
64
Page 80
components into account, this sightline exhibits high negative–velocity H I emission
(Complex C), Ovi absorption at a similar velocity, and broad, possibly redshifted
Ovii absorption. If Complex C is indeed nearby WHIM gas that has cooled and is
falling onto the Galaxy (e.g. Miville–Deschenes et al. 2005), then its presence could
indicate the presence of a large–scale WHIM filament in the same direction. In
this interpretation, the large nonthermal Doppler parameter of the Ovii absorption
could be a result of velocity shear, due either to the Hubble expansion over a
scale of ∼ 3 Mpc (with a corresponding density of 10−5 − 10−4.5 cm−3, assuming
b = 200 km s−1, pure Hubble broadening and a metallicity of 0.1 − 0.3× solar) or
the natural velocity distribution expected from infalling gas, or a combination of
both. The negative–velocity H I would then be gas that has “broken off” from the
filament and is now falling onto the Galaxy, with the Ovi at the same velocity
representing the cooling component of the gas. The velocity of the local standard of
rest is approximately perpendicular to the CMB rest frame in this direction, so no
significant additional velocity shifts are expected in an IGM scenario.
Such a picture is consistent with the general picture of galaxy formation and
accretion of gas onto galaxies from the IGM, and the temperatures and densities
inferred from the X-ray absorption are consistent with those expected from the
WHIM. However, unlike the Mrk 421 sightline, the simulations of Kravtsov et al.
(2002) do not predict high column densities of Ovii in this direction (though this
may be due to the limited resolution of the simulations). Furthermore, aside from
65
Page 81
Complex C there are no known structures in this direction that might indicate the
presence of a local filament — Wakker et al. (2003) note that the Canes Venatici
Galaxy Grouping is centered on this sightline at v ≈ 2400 km s−1, but this is far
higher than the velocity of the Ovii absorption.
Given these caveats and the large uncertainties on the X-ray measurements
(indeed, only one z = 0 absorption line has been strongly detected), the absorption
could also originate locally in hot Galactic halo or coronal gas. As discussed
previously, such an origin would require the Ovii to be a completely separate
component from any of the other observed components (H I, low– and high–velocity
Ovi) along this line of sight. Of course, there is also the possibility that the Ovii
absorption actually consists of multiple unresolved components, in which case a
multiphase solution may reconcile the discrepancy with the Ovi absorption. All line
ratio calculations were performed under the assumption of ionization equilibrium,
so nonequilibrium scenarios could provide substantively different predictions as
well. More detailed modeling and simulations of both the Galactic and IGM gas
distributions will be necessary to determine which scenario is most likely, and most
consistent with the data.
66
Page 82
3.4. Conclusions
Long–duration Chandra grating observations of the bright AGN Mrk 279 reveal
the presence of strong Ovii Kα absorption at a redshift consistent with zero. A
FUSE spectrum of the same source shows several additional Ovi components at
velocities near zero. Through kinematic, curve–of–growth, and ionization balance
modeling, we conclude the following:
1. A direct χ2 analysis of the Chandra spectrum coupled with absorption
line models constrains the Doppler parameter of the Ovii absorption to
be b > 74 km s−1 and b < 24 km s−1. This latter range is unlikely due to
the extremely high Ovii column densities required to produce the strong
absorption feature.
2. The Ovii Doppler parameter limits are inconsistent with the measured b values
for any of the v ≈ 0 Ovi absorption components. Additionally, the centroid
of the Ovii Kα line is inconsistent (at the 2.5σ level) with that of the OviHV,
indicating that the Ovii is not associated with any local Ovi component.
3. If the Ovii absorption is associated with a broad, undetected Ovi absorption
line, then a large Doppler parameter (b ≈ 200 km s−1) is required to provide
a single–phase solution for the Ovi, Ovii, and Oviii column densities. This
large value of b could be a result of either microturbulence, velocity shear from
67
Page 83
infalling gas, or broadening due to the Hubble expansion over a path length
of a few Mpc. If the line is purely Hubble–broadened, at b = 200 km s−1 a
pathlength of 3 Mpc and density of log n ≈ −5 is implied (assuming an oxygen
abundance of 0.3 times solar).
4. The large velocity dispersion, possible redshift, and lack of association with any
Galactic absorption components (as well as the proximity of HVC Complex
C) indicates that this X-ray absorption may be from a large–scale nearby
WHIM filament; however, a Galactic corona origin cannot be ruled out with
the current data.
68
Page 84
ID λresta λobs
b ∆vFWHM vobs Wλc log Ni
c,d Note
(A) (A) ( km s−1) ( km s−1) (mA)
X-ray:
O VII Kα 21.602 21.619 ± 0.009 600+400−600 236 ± 125 26.6 ± 6.2 16.19 ± 0.19 1
O VII Kβ 18.629 18.629 · · · · · · < 7.0 < 16.24
O VIII Kα 18.969 18.969 · · · · · · < 6.5 < 15.72
N VII Kα 24.781 24.781 · · · · · · < 7.0 < 15.51
Ne IX Kα 13.447 13.447 · · · · · · < 7.3 < 15.88
UV:
OviB 1031.926 1031.75 ± 0.01 112.3 ± 5.8 −50.8 ± 3.5 169.9 ± 5.5 14.21 ± 0.02
OviN 1031.926 1031.95 ± 0.01 69.9 ± 4.4 6.7 ± 1.8 92.6 ± 3.9 13.93 ± 0.02
OviHV 1031.926 1031.38 ± 0.01 53.3 ± 7.6 −159.9 ± 2.6 43.6 ± 3.5 13.58 ± 0.04
aRest wavelengths taken from Verner et al. (1996).
bIn the cases where upper limits were found, the line positions were frozen to the rest wavelengths.
cError bars are 1σ; upper limits are 2σ.
dColumn densities for X-ray lines are calculated assuming b = 200 km s−1; for UV lines the measured
b values are used.
Note. — (1) The column density given here for the Ovii Kα line is from the equivalent width
measurement assuming b = 200 km s−1, not the χ2 method described in §3.2.1.
Table 3.1. Observed z ≈ 0 absorption lines
69
Page 85
Fig. 3.1.— 10–47 A region of the coadded Chandra HRC–S/LETG spectrum of
Mrk 279 (top panel, black line) with the best–fit continuum model shown as the
red line, and residuals (bottom panel). The Ovii wavelength is marked by the solid
green line, and dotted green lines show the positions of the measured upper limits
listed in Table 1.
70
Page 86
Fig. 3.2.— 18-22 A region of Figure 3.1 showing in detail the z = 0 Ovii Kα/Kβ
and Oviii Kα regions. Here, the Ovii Kβ amplitude is constrained to the minimum
(unsaturated) value, Wλ(Kβ) = 0.156 × Wλ(Kα).
71
Page 87
Fig. 3.3.— Velocity relative to the local standard of rest of the local Ovii λ21.6
(top panel) and Ovi λ1032 absorption lines, with the best-fit model plotted in each
as the solid line; a representative error bar for the FUSE data points is shown at
left. Note the difference in scale between the two plots. A single–Gaussian fit to the
low–velocity Ovi absorption is also shown as the dotted line. The Ovii velocity is
inconsistent with that of the OviHV at the ∼ 2.8σ level.
72
Page 88
Fig. 3.4.— Allowed values of NOVII and b given the measured Ovii Kα equivalent
width and 1σ errors (shaded region) and Ovii Kβ 2σ upper limit (dashed line).
Values of NOVII and b for which the two measurements are consistent (within 95%
confidence) are denoted by the hatched region.
73
Page 89
Fig. 3.5.— Contours of constant equivalent width (solid) and FWHM (dashed) for
the low–velocity Ovi absorption, at the 1σ level. Red contours are derived from the
narrow low–velocity component and green contours from the broad component. The
inferred 1σ values of NOVI and b, as listed in Table 3.1, are shown as crosses.
74
Page 90
Fig. 3.6.— Regions of consistency in the temperature–density plane for the 2σ
Oviii/Ovii and Ovi/Ovii column density ratio upper limits. Here the Ovi upper
limit is calculated from a putative Ovi absorption line with b = 100 km s−1 superposed
on the Galactic Ovi absorption. While a consistent solution cannot be found for low
velocity dispersions, at higher values of b the contours begin to overlap.
75
Page 91
Fig. 3.7.— Same as Figure 3.6, but for a putative Ovi absorption line with
b = 200 km s−1.
76
Page 92
Chapter 4
The PKS 2155–304 Sightline
This chapter presents a detailed analysis of the Chandra Low–Energy
Transmission Grating observations of PKS 2155–304, a bright BL Lac object at
z = 0.116. Subsets of these data were studied by Fang et al. (2002) (searching
for redshifted absorption) and Nicastro et al. (2002) (z = 0 absorption), and
here we analyze both absorption systems using all available data. This coadded
spectrum represents the second highest–quality Chandra grating observation of an
extragalactic source in terms of counts per resolution element (after Mrk 421), and
the last sightline for which potentially extragalactic, z ≈ 0 X-ray absorption lines are
likely to be detected (excluding 3C 273, which lies in the direction of a supernova
remnant). Since the physical properties of the z = 0 absorption toward Mrk 421 and
Mrk 279 differed substantially in some ways (velocity dispersion and offset from the
Ovi HVCs), a third sightline provides valuable insight into the global properties of
77
Page 93
this absorption. Additionally, the relatively high redshift of PKS 2155–304 provides
a large path length over which to find intervening WHIM absorption systems.
4.1. Data Reduction and Measurements
4.1.1. Chandra
PKS 2155–304 has been observed numerous times for calibration and other
purposes with all possible combinations of the Chandra gratings and detectors. As
the strongest absorption lines previously observed have been lines from C, N, and
O at λ ∼> 18 A, for the purposes of this study we only include data from the Low
Energy Transmission Grating (LETG) since it has the highest effective area in this
wavelength regime. Chandra’s two X-ray cameras, the High Resolution Camera
(HRC) and Advanced CCD Imaging Camera (ACIS), each include separate detector
arrays for imaging (I) and grating spectroscopy (S). Although LETG observations
taken with HRC–I and ACIS–I are available in the archive, their calibration is less
certain and wavelength range more restricted than those of the spectroscopic arrays,
and so they are excluded from this analysis.
The remaining datasets include 8 employing the HRC–S/LETG instruments
and 24 with ACIS–S/LETG. Of these latter observations, however, 15 have large
pointing offsets (typically 6′ − 14′), presumably intended to characterize the off–axis
line spread function and effective area. Since the spectral resolution degrades
78
Page 94
significantly at these large offsets, only the nine ACIS–S observations with |∆θ| ≤ 1.′5
are considered here. The resulting 17 observations, listed in Table 4.1, contain a total
of 483 ks of exposure time and ∼ 2100 counts per 0.05 A resolution element (CPRE)
at 21.5 A in the first LETG spectral order (253.7 ks and 880 CPRE in ACIS–S;
229.3 ks and 1200 CPRE in HRC–S). In theory this should provide roughly half the
signal–to–noise ratio obtained in the extremely high–quality LETG spectrum of Mrk
421 outburst (6000 CPRE; Nicastro et al. 2005a; Williams et al. 2005).
All datasets were fully reprocessed using the Chandra Interactive Analysis
of Observations (CIAO) software, version 3.3, with the corresponding Calibration
Database (CALDB) version 3.2.11. This CALDB version includes models for the
ACIS–S time–dependent quantum efficiency degradation as well as preliminary
corrections to the nonlinearities in the HRC–S/LETG wavelength scale. First–order
spectra were then extracted, and response matrices built, using the standard
CIAO routines. LETG spectral orders cannot be separated with HRC–S due to
this detector’s intrinsic lack of energy resolution, so the resulting spectrum is an
overlapping superposition of all orders. We thus built all HRC–S/LETG response
matrices for orders −6 to +6; past experience (e.g., Williams et al. 2006a) has shown
that this is sufficient to accurately model higher–order contamination.
The individual spectra from each instrument were then coadded, both to
allow searches by eye for intervening absorption lines and to make it easier to
1See cxc.harvard.edu/ciao/ and cxc.harvard.edu/caldb/
79
Page 95
assess the goodness of fits. First, the positive and negative spectral orders from
each observation (and their corresponding response matrices) were coadded. As
PKS 2155–304 is highly variable in the X-ray band, we performed quick fits to
determine the flux near 21 A at the time of each observation. The response matrices
were then weighted by a factor of fλ(21A) × texp and coadded; these fluxes and
weights are also listed in Table 4.1. Note that, as a result of the dithering strategy
employed during observations, Chandra/LETG lacks the narrow chip–gaps and other
detector features seen in XMM–Newton grating spectra (Williams et al. 2006b);
weighting the response matrices before coaddition substantially reduces broad
residuals across the LETG band, but is not essential for narrow absorption line
measurements.
The resulting spectra were fit using the CIAO Sherpa utility. Simple powerlaw
continua (with foreground Galactic absorption as a free parameter) were fit
independently for the ACIS and HRC spectra over the 10–47 A wavelength range.
To improve the consistency of the fit near elemental edges, the foreground absorber
abundances of carbon, nitrogen, oxygen, and neon were allowed to vary. The
resulting best–fit powerlaw slopes are similar (Γ = −0.63 and −0.45 for ACIS and
HRC respectively, where fλ ≈ λ−Γ), but the Galactic absorption and abundances
vary somewhat between the two instruments, perhaps due to calibration uncertainties
or a degeneracy between Γ and NH over this restricted wavelength range. The
continuum fits for ACIS and HRC are shown in Figures 4.1 and 4.2 respectively.
80
Page 96
Several absorption lines, including the Ovii, Oviii, and Ne ix Kα transitions
at z = 0, are immediately visible. These lines were modeled with narrow
(FWHM< 50 mA) Gaussian features added to the fitted continua. Since there may
be some lingering systematic uncertainties in the HRC–S/LETG wavelength scale
even with the new correction routines, wavelengths and strengths of absorption
features were first allowed to vary independently for the spectrum produced by each
instrument. Each line’s equivalent width was then determined using a joint fit to
the ACIS and HRC spectra with the requirement that the equivalent width match
between the two instruments, i.e. the normalized Gaussian line amplitudes AHRC
and AACIS (where A corresponds to the integral of the Gaussian, not the height)
were fixed according to:
fλ,HRCAHRC = fλ,ACISAACIS (4.1)
Wavelength and equivalent width errors were determined for this joint fit using
the “projection” command in Sherpa, allowing the HRC and ACIS continuum
normalizations to vary. These quantities for all measured z = 0 lines (as well as
upper limits on Ne ix and Ovii Kγ) are reported in Table 4.2, and Figures 4.3 and
4.4 show the best–fit models for all detected lines in ACIS and HRC respectively.
81
Page 97
4.1.2. FUSE
The reduction and analysis of the PKS 2155–304 FUSE data were performed in
a manner nearly identical to that described for the Mrk 421 (Williams et al. 2005)
and Mrk 279 (Williams et al. 2006a) sightlines; a brief summary follows. Three
observations of PKS 2155–304totaling 120 ks were available in the FUSE section of
the Multimission Archive at STScI website.2 The calibrated data were downloaded
and individual exposures cross-correlated over the 1030–1040 A range, where many
strong absorption lines are present, to compensate for small (typically < 10 mA)
variations in the wavelength scale. The cross–correlated spectra for each observation
were then coadded, and the resulting spectra in turn cross–correlated and coadded
to produce a final coadded FUSE spectrum. The absolute wavelength scale was
checked by comparing the positions of the strong, narrow Si II λ1020.699 and Ar I
λ1048.220 absorption lines to the H I 21 cm emission. Wakker et al. (2003) find
that essentially all of the H I is concentrated in a single Gaussian component at
−4 km s−1, which matches quite well the measured Ar I and Si II velocities (−5.4 and
−5.9 km s−1 respectively, with about 0.6 km s−1statistical error). Since there may
be small (∼ few km s−1) systematic errors arising in the H I measurement and/or
the physical relation between H I and the two FUSE–measured species, we will thus
assume the FUSE wavelength calibration is correct for the purposes of this work.
2archive.stsci.edu/
82
Page 98
The FUSE spectrum shows strong Ovi λ1032 absorption at v ≈ 0 as well
as two distinct high–negative velocity Ovi components (hereafter referred to as
OviHVC1 and OviHVC2 in order of increasing absolute velocity). The 1029–1034 A
region of the spectrum was fit with a constant continuum plus a single Gaussian for
each of the three Ovi components. However, this provided a poor fit for the strong
low–velocity Ovi component so another Gaussian was added at v ≈ 0 to improve
the fit. Figure 4.5 shows the resulting data and best–fit model, and the parameters
of the four Gaussian components are listed in Table 4.2. Note that the velocity of
OviHVC2 component is inconsistent with the z = 0 Ovii Kα velocity at the ∼ 3σ
level assuming the statistical error on the line measurement, or 2.6σ if the nominal
systematic wavelength uncertainty of 10 mA is adopted3, indicating that the Ovii
and OviHVC2 components may be kinematically distinct. However, since wavelength
scale errors in Chandra are still not well–determined, this should not be considered
a firm result.
The other Ovi doublet line at 1037.6 A is also visible in the spectrum, and
in principle can be useful for curve–of–growth diagnostics when the Ovi λ1032
line is saturated. With the high resolution of FUSE (λ/∆λ ≈ 15000), however,
the 1032 A line’s shape and strength has in the past been quite sufficient for these
measurements. Furthermore, the 1037 A line components (particularly the HVCs)
are heavily blended with nearby Galactic interstellar medium lines such as C II∗.
3See cxc.harvard.edu/cal/
83
Page 99
Since this blending can introduce additional systematic error and even slight Ovi
saturation appears to be rare (e.g. Wakker et al. 2003), we will disregard the 1037 A
Ovi line in this analysis.
4.2. Analysis
4.2.1. Doppler Parameters and Column Densities
The low resolution of the Chandra gratings compared to UV and optical
spectrographs presents unique challenges for column density measurements, since
essentially all non–quasar absorption lines are far narrower than the 50 mA
(∼ 700 km s−1 at 20 A) LETG line–spread function. The lack of line width
information prevents direct measurement of the profile shape, and hence the degree
of saturation for any given line cannot be directly determined. If multiple absorption
lines from the same ionic species are detected, however, the relative equivalent
widths of these lines can instead be used to place limits on the column density
(NOVII) and velocity dispersion (or Doppler parameter, b) of the medium.
In the case of PKS 2155–304, the Ovii Kα and Kβ lines are strongly detected,
and an upper limit is measured for the Kγ line. If all these lines were unsaturated,
the equivalent widths would scale as Wλ ≈ fijλ2 where fij is the absorption oscillator
strength. Saturation effectively decreases the equivalent widths of strong (high–fij)
84
Page 100
lines while leaving weaker lines in the series more or less unaffected; thus, the ratio
Wλ(Kβ)/Wλ(Kα) increases with respect to the simple (unsaturated) proportionality
above. For Ovii, the expected Kβ/Kα equivalent width ratio is 0.156, while we
measure a ratio of 0.36±0.12, indicating that the Ovii Kα may be slightly saturated
(though it is also consistent with no saturation at the ∼ 2σ level).
To place more quantitative constraints on NOVII and b, we employ the technique
used in Williams et al. (2006a) for Mrk 279. For a grid of points in the NOVII − b
plane, equivalent widths and apparent line widths were calculated numerically (using
Voigt absorption line profiles) for the Ovii Kα, Kβ, and Kγ transitions. These lines
were then added to the continuum model in Sherpa, and χ2 calculated with the
“goodness” command, for every value of NOVII and b. Figure 4.6 shows the contours
of 1σ, 2σ, and 3σ confidence calculated in this manner.
As this figure shows, the minimum χ2 is found at b = 52 km s−1 and
log(NOVII) = 16.2, with the 1σ confidence region stretching between b = 35−94 km s−1
and log(NOVII) = 15.9 − 16.5. Additionally, another 1σ region can be found at
b < 19 km s−1 with a higher column density (log(NOVII) ≈ 17.5) required to produce
consistency with the spectrum. Such high Ovii column densities are unlikely to
be produced in a cold (Tmax ∼< 3 × 105 K), weakly photoionized medium without
producing large amounts of narrow Ovi absorption, so a low–b solution appears
unlikely. However, it is important to note that no value of b can be ruled out at the
2σ confidence level from this curve–of–growth analysis alone; as mentioned above,
85
Page 101
the absorption is consistent with a completely unsaturated medium at this level,
and lower–b, higher–NOVII solutions are also possible in the regions demarcated by
the 2σ and 3σ contours in Figure 4.6. Column densities for all X-ray lines (listed in
Table 4.2) are calculated by assuming the χ2–minimizing b value above. While most
of the lines are too weak for this choice to make a significant difference in the Nion
determination, it should be kept in mind that the systematic uncertainty in NOVII
may be larger than the statistical errorbars.
Determination of these parameters for the UV Ovi λ1032 absorption is
decidedly more straightforward since the lines are fully resolved by FUSE. Since
saturation can make absorption lines broader than would be expected just from the
Doppler parameter of the gas, the measured line width cannot be used directly as a
surrogate for b. Instead, we calculate apparent Ovi FWHM values and equivalent
widths over a grid of NOVI and b, and find the regions within this grid that are
consistent with the measured ∆vFWHM and Wλ values.
Figure 4.7 shows these tracks for the low–velocity Ovi components. Contours
of constant ∆vFWHM are roughly vertical while constant Wλ are horizontal in
the unsaturated regime. In this case both Ovi1 and Ovi2 appear to be at most
weakly saturated, so the ∆vFWHM and Wλ contours overlap nearly orthogonally,
producing tight constraints on both parameters for both components. We find
that b = 41.5 ± 2.5 km s−1 and b = 51 ± 5 km s−1 for Ovi1 and Ovi2 respectively,
with column densities of log(NOVI) = 14.06 ± 0.02 and 13.94 ± 0.02. These Ovi
86
Page 102
b values are fully consistent with the Ovii 1σ measurement. Since the Ovi
low–velocity lines are essentially unsaturated, and the HVCs are weaker still
(but with comparable apparent line widths), we can safely assume that they are
unsaturated; thus, we assume their b–values are directly calculable from their
FWHMs (i.e. FWHM = 2b√
ln 2 = 1.665b).
4.2.2. Temperature and Density Diagnostics
With estimates for ionic column densities, constraints on the temperature and
density of the absorbing medium can be derived. Although collisional ionization is
expected to be the dominant physical process in either the extended local WHIM or
a hot Galactic corona, photoionization from the extragalactic UV/X-ray background
is expected to significantly alter the ionization balance of the low–density WHIM (cf.
Nicastro et al. 2002). To find the most general set of conditions which can produce
the observed highly–ionized ion ratios, both collisional and photo–ionization must
be considered.
For this sightline we follow the same analysis we employed for Mrk 279 in
Williams et al. (2006a). Assuming a fixed z = 0 metagalactic ionizing background
model from Sternberg et al. (2002), the ionization parameter U = nγ(E > 13.6eV)/ne
is simply the inverse of the electron density. The ionization balance code Cloudy
(version 05.04; Ferland et al. 1998) was employed to calculate relative abundances
87
Page 103
of all measured ions over a range of T = 104.5 − 107.4 K and ne = 10−7 − 1 cm−3 (or
U = 100.7 − 10−6.3) with grid spacings of 0.1 dex in each quantity, fully encompassing
the range of temperatures and densities expected in WHIM and Galactic corona
models.
With a grid of Ni computed as a function of temperature and density, the
problem can be inverted to determine which sets of T and ne are consistent with
the measured ionic column densities. However, since the local X-ray absorption is
produced in gas too hot to be detectable in neutral hydrogen emission (and Lyα
absorption at z ≈ 0 is invariably wiped out by the local interstellar medium damping
wing), no information on the overall metallicity can be derived from the data. Thus,
it is more useful to find the log T − log ne regions defined by column density ratios.
Since Ovii is by far the best–measured ion that unambiguously arises in local
warm–hot gas, we calculate all other ion column density ratios relative to NOVII.
Column density ratios of different ions of the same element are independent
of metallicity, and so depend only on the physical state of the medium. Thus,
if the Oviii and Ovii absorption arise in the same gas phase, the NOVIII/NOVII
ratio provides the most rigorous constraints on the temperature and density of the
warm–hot gas. Likewise, if any one of the four measured Ovi components exists
in this same phase, the NOVI/NOVII ratio should be consistent with an overlapping
set of temperatures and densities. Figure 4.8 shows the 2σ constraints derived
from NOVIII/NOVII and NOVI/NOVII for each of the Ovi components. Note that
88
Page 104
at high densities (ne ∼> 10−4 cm−3) the temperature constraints are essentially
constant, but at lower densities photoionization becomes significant and a lower
temperature is sufficient to produce the same column density ratios. The log T
and log ne values derived from the measured NOVIII/NOVII are consistent with the
NOVI/NOVII constraints for all four components, with a typical minimum density of
ne ∼> 10−5 cm−3. Thus, association of the X-ray oxygen absorption lines with any
one of the UV Ovi components cannot be ruled out for this sightline. If the medium
is collisionally ionized, the NOVIII/NOVII ratio provides strong 2σ temperature
constraints (assuming a Doppler parameter b ≈ 52 km s−1) of 6.18 < log T < 6.36.
Similar constraints can be derived from other measured X-ray lines, though
since the ratio NCVI/NOVII depends on the relative [C/O] abundance (for example),
these constraints are more prone to systematics than those solely employing one
atomic species. Moreover, since the Ovii Doppler parameter cannot be pinned down
to better than 1σ, NOVII can vary with different b values thus significantly altering
the measured column density ratios. For the low–b solution (b < 19 km s−1), the
measured NOVIII/NOVII requires temperatures of log T ≈ 6.2. On the other hand,
such a low velocity dispersion implies a maximum temperature of log Tmax = 5.5 so
the low–b solution does not appear to be physically possible. For these X-ray line
diagnostics, we thus consider only the best–fit 1σ region of b = 52+42−35 km s−1 and a
large–velocity dispersion, low saturation (b ≈ 200 km s−1) solution.
89
Page 105
Figures 4.9 and 4.10 show the constraints, for b = 52 and b = 200 respectively,
derived from the ratios of Ne ix, Cvi, and Oviii to Ovii. The column density
limits measured for Nvi and Nvii did not provide any useful constraints (i.e., they
were consistent with nearly the entire range of temperatures and densities) and were
excluded from these figures for clarity. Furthermore, the Cv ion is expected to form
in cooler gas than that producing the Ovii and Oviii absorption (most likely the
local warm ISM and/or the Galactic thick disk), and so the NCV/NOVII constraint
was also not considered for this analysis.
In the best–fit b plot (Figure 4.9), the constraints derived from all three
column density ratios overlap quite well in the collisionally–ionized density regime
(ne ∼> 10−5 cm−3). However, at a large velocity disperion (b = 200; Figure 4.10),
there is essentially no set of log T and log ne for which the three constraints overlap.
The Cvi and Ne ix lines are relatively weak and not as affected by saturation as
the Ovii, so this change is driven primarily by the decrease in NOVII at higher
velocity dispersion. It should be re–emphasized that the contours derived from NCVI
and NNeIX depend on [C/O] and [Ne/O]; however, if the abundance mixture of this
absorber is roughly solar, then the measured column densities indicate that the
previously derived Doppler parameter range (b = 35 − 94 km s−1) fit the data better
than a high–b, unsaturated medium.
90
Page 106
4.2.3. z = 0.055 Absorption Reported by Fang et al.
In a previous study incorporating a subset of the data analyzed herein
(ACIS–S/LETG observation IDs 1703, 2335, and 3168), Fang et al. (2002, hereafter
F02) report the detection of an absorption line at 20.02 ± 0.015 km s−1 with
equivalent width 14.0+7.3−5.6 mA, possibly corresponding to Oviii at a velocity of
16634±237 km s−1 (z = 0.055±0.001). Several previously discovered Lyα absorption
lines and a small cluster of H I galaxies appear at a similar velocity in this direction,
so such an absorber may be indicative of intragroup medium or an associated
large–scale WHIM filament.
This absorption line is clearly visible in our coadded ACIS spectrum, but is
not visible in the HRC spectrum (Figure 4.11). A fit to the line in ACIS yields a
wavelength of 20.03 ± 0.01 A and an equivalent width Wλ = 7.5 ± 2.1 mA, both
consistent with the F02 measurement. An upper limit for a line at this wavelength
(±0.02 A, to account for possible discrepancies in the HRC/LETG wavelength
scale) was calculated and found to be Wλ < 12.5 mA (2σ confidence), so the
ACIS measurement is not ruled out by the HRC detection. Moreover, our best–fit
equivalent width is roughly half that (and hence half the Oviii column density)
measured by F02. If real, this detection would still be fully consistent (albeit with
large errors) with the predicted number of Oviii absorbers per unit redshift (F02,
Figure 2).
91
Page 107
However, the existence of this line is brought into question by the higher–quality
ACIS/LETG spectrum of Mrk 421 (see Figure 4.11). In this spectrum, a weak
detector feature is seen at ∼ 19.97 A, or 50 mA from the wavelength of the feature
seen toward PKS 2155–304. The instrument model of the Mrk 421 spectrum appears
to partially compensate for this detector feature, but no such feature is seen in the
instrumental response of the PKS 2155–304 spectrum. The presence of a detector
feature near this wavelength in the Mrk 421 spectrum is worrisome, as it may be
indicative of a transient anomaly (or one that was deemed insignificant, and removed
from the instrument model before the bulk of the PKS 2155–304 observations).
Because of this, coupled with the apparent absence of other highly–ionized absorption
lines at this velocity in the new higher–quality PKS 2155–304 data, we conclude
that there is a high likelihood that the previous detection was due to a feature in the
ACIS-S/LETG system.
4.3. Discussion
4.3.1. Comparison to Other Lines of Sight
PKS 2155–304 represents the third and final Chandra data set we have analyzed
for which relatively weak (∼ 10 mA) absorption, uncontaminated by known hot
foreground gas (as in 3C 273, situated along the line of sight to a supernova remnant)
can be detected at high confidence. Although the number of such sightlines is small,
92
Page 108
similarities and differences in the detected absorption are already beginning to
emerge.
Mrk 421
The Chandra LETG spectrum of Mrk 421 contains is the highest–quality
grating spectrum of an extragalactic source to date, with roughly three times the
counts per resolution element at 21 A as the PKS 2155–304 data analyzed here
(Nicastro et al. 2005a; Williams et al. 2005). Many of the ionic species seen toward
Mrk 421 (particularly Ovii, Oviii, Ne ix, and Cvi) were also detected in the
PKS 2155–304 spectrum, allowing direct comparisons to be made. In particular,
the Ovii absorption (the strongest–detected ion in both cases) shows strikingly
similar properties between the two objects–NOVII = 16.23 ± 0.21 in Mrk 421
versus 16.09 ± 0.19 in PKS 2155–304. Both Ovii absorbers exhibit consistent
velocity dispersions as well, with bOVII = 24 − 55 km s−1 (2σ limits) in Mrk 421 and
35− 94 km s−1 in PKS 2155–304 (though this latter quantity is the 1σ range, and 2σ
limits could not be determined).
One of the most surprising features of the Mrk 421 Chandra spectrum is the
presence of a weak (2.0 mA) absorption line at 22.02 A, the expected wavelength of
the Ovi Kα inner–shell transition. Even though both this transition and the 1032 A
lines should both trace the Ovi ground state, the column density derived from the
observed Kα line is a factor of ∼ 4 higher than that derived from the UV transition.
93
Page 109
If this Kα transition is a better tracer of the true Ovi column density than the
UV line, then the Ovii toward Mrk 421 almost certainly arises in a low–density,
photoionized medium.
Although the Ovi λ1032 absorption strength toward PKS 2155–304 is
comparable to the Mrk 421 sightline, unfortunately the Chandra spectrum does
not have sufficient signal–to–noise to detect the Kα line. Thus, for PKS 2155–304
we cannot determine if there is the same discrepancy between the UV and X-ray
Ovi, or if the inferred Ovi Kα column density also implies a photoionized medium
along this sightline. If the Ovi Kα measurement is disregarded, then the Mrk
421 spectrum implies slightly lower–but marginally consistent–temperatures in the
collisionally ionized regime (log T = 6.1 − 6.2; 2σ limits from the Oviii/Ovii ratio)
than PKS 2155–304 (log T = 6.18 − 6.36). The lower density limit from Mrk 421
is somewhat more stringent than that derived from the PKS 2155–304 oxygen ion
ratios (log ne > −4.7 versus > −5.5), though this may again be primarily an issue of
spectral quality.
Mrk 279
The z = 0.03 Seyfert galaxy Mrk 279 is significantly less luminous than either
PKS 2155–304 or Mrk 421, but it was observed sufficiently long with Chandra
HRC–S/LETG to produce a reasonably high–quality spectrum in which strong
z = 0 Ovii Kα absorption was detected (Williams et al. 2006a). Two features of
94
Page 110
this absorption were particularly interesting: (1) the unusually strong Ovii Kα
absorption (Wλ = 26.6 ± 6.2 mA), coupled with a tight upper limit on the Kβ line,
indicated that the absorption was best described as an unsaturated medium (with
a 2σ lower limit of b > 77 km s−1 on the Doppler parameter); and (2) the Ovii
absorption appears slightly redshifted, making its velocity inconsistent with the Ovi
HVC’s negative velocity at the 2.5σ level.
As previously mentioned, the velocity of the Ovii toward PKS 2155–304
appears to be inconsistent with the OviHVC2 velocity at the 2.6σ level, assuming
the nominal HRC–S/LETG wavelength scale uncertainty of 10 mA. Until systematic
errors in the wavelength scale can be better understood, however, this should not be
taken as a definitive result. Moreover, while the Ovii toward Mrk 279 could not
be directly associated with either low–velocity Ovi component since their velocity
dispersions were significantly different, such an association cannot be ruled out in
the PKS 2155–304 absorption: all of the Ovi Doppler parameters fall within the 1σ
b confidence interval found for the Ovii absorption.
Temperature and density constraints on the z = 0 absorption toward Mrk 279
and PKS 2155–304 are consistent with each other, though this is not surprising–since
the quality of the Mrk 279 spectrum is lower, only an upper limit on temperature
and a lower limit on density could be derived.
95
Page 111
4.3.2. Where is the Absorption?
In principle, the degree of photoionization of an absorbing medium (and hence
an estimate of the gas density) can be derived from ionic column density ratios
(§4.2.2). However, in the case of PKS 2155–304, the errors are large enough that no
upper limit on the gas density can be found, i.e. it is fully consistent with collisional
ionization. Depending on the assumptions made (in particular, which if any of the
Ovi components are associated with the Ovii and Oviii), the minimum density
of this medium appears to be log ne ∼> −5.5. The best–fit Ovii column density is
log NOVII = 16.09 and Ovii is by far the dominant ionization state in this medium.
If the gas has a metallicity of 0.3× solar (comparable to that observed in the diffuse
intracluster medium), then the total hydrogen density is roughly log NH ≈ 19.9 and
the thickness of the absorber d ∼< 1025.4 cm = 8.4 Mpc.
Thus, under a set of reasonable assumptions, the observed absorption is
consistent with an extended extragalactic medium, but within the very large errors
it is just as reasonable to associate it with a local hot Galactic corona. Although an
association between the X-ray absorber and low–velocity Ovi cannot be ruled out
from these data alone, the properties of the X-ray absorber are quite similar (again
within the errors) to those studied along the Mrk 421 and Mrk 279 sightlines. These
latter two X-ray absorbers are definitely not associated with the low–velocity Ovi
absorption arising in the Galactic thick disk, so if they indicate the presence of an
96
Page 112
additional hot Galactic component, then the derived properties of the PKS 2155–304
absorber are not in conflict with the other measurements of this component.
4.3.3. Comparison to Nicastro et al. (2002)
In their study of three HRC–S/LETG observations of PKS 2155–304
(observation IDs 331, 1013, and 1704), Nicastro et al. (2002, hereafter N02) also
detected OviiKα and Kβ, Oviii, and Ne ix, albeit at lower confidence. They
also analyze FUSE data of the same sightline, but at that time only 39 ks were
available, or about one–third of the exposure time analyzed here. As it turns out,
the addition of new Chandra and FUSE data brings about significant changes in the
interpretation of the local absorption, in two important ways.
First, while N02 were able to fit the observed Ovi λ1032 line with two Gaussian
components (one low–velocity narrow line and a broader, blueshifted HVC), the
new higher–quality spectrum reveals that the “broad” component is actually two
distinct HVCs, and the narrow low–velocity Ovi is best fit with two components.
Second, while the Ovii Kβ line was previously not detected strongly enough to
place constraints on the Doppler parameter of the absorption, here the Kα and Kβ
lines hint at some degree of saturation and so the Ovii column density we employ
in our analysis is about 0.5 dex higher than that reported by N02. Both of these
effects cause a sharp decrease in the NOVI/NOVII ratio, which in turn removes the
97
Page 113
need for a photoionization contribution. Indeed, when we calculate temperature
and density constraints assuming b = 200 (as N02 had done; see Figure 4.10) we
also find that a high–density, collisionally–ionized solution cannot be found without
modifications to the relative abundances. This highlights the major improvements
in diagnostic power that can be made by accumulating large numbers of counts per
resolution element, either through very long exposures or observations of especially
bright background sources.
4.4. Conclusions
Using all available Chandra LETG data on PKS 2155–304, we have analyzed in
detail the ionization and kinematic state of the warm–hot z ≈ 0 absorbing medium.
We find a Doppler parameter range of b = 35 − 94 km s−1 (1σ limits; 2σ limits could
not be found), which is consistent with the absorption seen toward both Mrk 421 and
Mrk 279 (though the best–fit value best matches the former sightline). Assuming
that the Doppler parameter lies in this range, ionic column densities of Ovii, Oviii,
Ne ix, and Cvi are consistent with collisional ionization at log T (K) = 6.18 − 6.36,
though a low–density photoionized WHIM cannot be ruled out. Unlike the other
two previously analyzed sightlines, the Ovii absorption toward PKS 2155–304 may
be associated with either one of the low–velocity Ovi λ1032 components seen in
FUSEor a high–velocity Ovi cloud at v = −130 km s−1 (though its velocity may be
98
Page 114
inconsistent at the ∼ 2.5σ level with another Ovi HVC at v = −234 km s−1). The
intervening Oviii absorber at z = 0.055 reported by Fang et al. (2002) is detected
in ACIS, but may be related to an unmodelled detector feature that is seen in other
ACIS-S/LETG spectra.
It is notable that the Chandra data for this line of sight hint at a number of
interesting results (particularly the possibly low Doppler parameter) but the data
are not quite of sufficient quality to confidently confirm them. PKS 2155–304 is
quite possibly the only other source bright enough to obtain a Chandra LETG
spectrum with ∼ 6000 counts per resolution element, comparable to the Mrk 421
spectrum analyzed by Williams et al. (2005), in a reasonable amount of time. Such
a spectrum would not only allow a direct comparison of the z ≈ 0 absorption along
two lines of sight, but would also provide a path length four times larger than Mrk
421 to search for “missing baryons” in intervening WHIM filaments. With two such
systems detected in Mrk 421, a correspondingly larger number could be detected in
a PKS 2155–304 spectrum at comparable signal–to–noise. Thus, this is a case where
longer observations of this object are both attainable and have the potential for
great scientific benefit.
99
Page 115
Obs ID Date texp fλ(21A)a Wib
(ks) (ks−1 cm−2 A−1)
ACIS–S/LETG
1703 2000 May 31 25.2 3.4 0.121
2335 2000 Dec 06 29.1 2.4 0.090
3168 2001 Nov 30 28.8 6.7 0.223
3668 2002 Jun 11 13.5 6.9 0.120
3707 2002 Nov 30 26.9 1.6 0.054
4416 2003 Dec 16 46.5 3.2 0.174
6090 2005 May 25 27.5 3.2 0.106
6091 2005 Sep 19 29.2 2.6 0.087
6927 2006 Apr 02 27.0 0.8 0.025
HRC–S/LETG
331 1999 Dec 25 62.7 9.7 0.514
1013 2001 Apr 06 26.6 2.9 0.065
1704 2000 May 31 25.8 3.7 0.081
3166 2001 Nov 30 29.8 8.4 0.212
3709 2002 Nov 30 13.7 2.3 0.026
4406 2002 Nov 30 13.9 2.4 0.028
5172 2004 Nov 22 26.9 1.8 0.041
6923 2006 May 01 29.9 1.3 0.032
aBackground–subtracted photon flux at 21A; the HRC–S
values are apparent fluxes including all higher orders.
bWeight factors for coadding the response matrices, calculated
as Wi = fλ,i(21A)texp,i/Σi(fλ,i(21A)texp,i).
Table 4.1. Chandra observation log
100
Page 116
ID λresta λobs
b ∆vFWHM vobs Wλc log Ni
c,d Note
(A) (A) ( km s−1) ( km s−1) (mA)
X-ray:
C V Kα 40.268 40.227+?−.015 · · · −305+?
−112 11.4 ± 5.1 15.22+0.26−0.33 1
C VI Kα 33.736 33.732+.011−.007 · · · −36+98
−62 5.6 ± 2.5 15.16+0.18−0.27
O VII Kα 21.602 21.611+.002−.008 · · · 125+28
−111 11.6 ± 1.6 16.09 ± 0.19
O VII Kβ 18.629 18.618 ± .007 · · · −177 ± 113 4.2 ± 1.3 16.09+0.17−0.21
O VIII Kα 18.969 18.987+.003−.008 · · · 285+47
−126 6.7 ± 1.4 15.80+0.11−0.13
Ne IX Kα 13.447 13.451+.010−.003 · · · 89+223
−67 4.5 ± 1.1 15.83 ± 0.21
N VI Kα 28.787 28.787 · · · · · · < 8.4 < 15.39
N VII Kα 24.781 24.781 · · · · · · < 5.0 < 15.39
O VI Kα 22.019 22.019 · · · · · · < 5.7 < 15.36
O VII Kγ 17.768 17.768 · · · · · · < 5.3 < 16.71
(cont’d)
Table 4.2. Observed z ≈ 0 absorption lines
101
Page 117
Table 4.2—Continued
ID λresta λobs
b ∆vFWHM vobs Wλc log Ni
c,d Note
(A) (A) ( km s−1) ( km s−1) (mA)
UV:
Ovi1 1031.926 1032.11 ± 0.01 76.4 ± 4.1 53.5 ± 2.9 120.5 ± 3.8 14.06 ± 0.02
Ovi2 1031.926 1031.85 ± 0.01 90.1 ± 8.1 −22.1 ± 2.9 98.2 ± 4.4 13.94 ± 0.02
OviHVC1 1031.926 1031.48 ± 0.01 74.2 ± 5.8 −129.7 ± 2.9 73.1 ± 4.0 13.81 ± 0.03
OviHVC2 1031.926 1031.12 ± 0.02 80.3 ± 14.5 −234.3 ± 5.8 46.6 ± 4.5 13.59 ± 0.05
aRest wavelengths taken from Verner et al. (1996), except Ovi Kα which is from the laboratory
measurements of Schmidt et al. (2004).
bIn the cases where upper limits were found, the line positions were allowed to vary within 20 mA
of the rest wavelengths. Measured wavelengths are taken from ACIS–S since its dispersion relation is
thought to have the fewest nonlinearities; however, only the statistical fit error is given in this column
(i.e., not taking into account possible wavelength scale uncertainties of up to ∼ 10 mA).
cError bars are 1σ; upper limits are 2σ.
dColumn densities for X-ray lines are calculated assuming the 1σ Doppler parameter region found in
Figure 4.6 (35-94 km s−1 for Ovii); for UV lines the measured b values are used.
Note. — (1) An upper error bar could not be formally determined for the wavelength of this line.
102
Page 118
Fig. 4.1.— Chandra grating over the 10–47A range from ACIS-S/LETG with residuals
from the best–fit continuum shown in the lower panel. Several narrow absorption lines
are apparent in these residuals.
103
Page 119
Fig. 4.2.— Same as Figure 4.1, showing the continuum fit for the Chandra HRC–
S/LETG spectrum.
104
Page 120
Fig. 4.3.— ACIS–S/LETG data (points) and best–fit model (histogram) near each
of the six z = 0 X-ray absorption lines detected toward PKS 2155–304.
105
Page 121
Fig. 4.4.— Detected z = 0 absorption lines in HRC–S; see Figure 4.3 for details.
106
Page 122
Fig. 4.5.— Ovi λ1032 region of the PKS 2155–304 FUSE spectrum. The four
Gaussian components used to model the z ≈ 0 Ovi complex are labeled.
107
Page 123
Fig. 4.6.— Ovii Doppler parameter and column density contours (at the 1σ, 2σ, and
3σ confidence levels) determined by simultaneously Ovii Kα, Kβ, and Kγ lines in
the joint LETG/ACIS+HRC Chandra spectrum.
108
Page 124
Fig. 4.7.— Column density and velocity dispersion constraints derived from Ovi
λ1032 equivalent widths (roughtly horizontal lines) and FWHMs (vertical lines),
for the Ovi1 (solid) and Ovi2 (dashed) components. Regions of overlap, marked
with crosses, denote the approximate 1σ confidence intervals on NOVI and b for each
component.
109
Page 125
Fig. 4.8.— Oxygen ion temperature and density constraints for NOVIII/NOVII and
NOVI/NOVII for each of the four measured Ovi components.
110
Page 126
Fig. 4.9.— Temperature and density constraints from ion column density ratios
relative to NOVII, where column densities are derived assuming the best–fit region
near b = 50 km s−1 shown in Figure 4.6.
111
Page 127
Fig. 4.10.— Same as Figure 4.9, but with column densities derived assuming minimal
saturation (b = 200 km s−1).
112
Page 128
Fig. 4.11.— ACIS–S and HRC–S/LETG (top and center panels, respectively) spectra
of PKS 2155–304 near the wavelength of the z = 0.055 Oviii intervening feature
reported by (Fang et al. 2002), and the same portion of the Mrk 421 ACIS–S/LETG
spectrum. The presence of a feature near this wavelength in both ACIS–S/LETG
spectra and not in the HRC spectrum indicates that this line is most likely a detector
feature.
113
Page 129
Chapter 5
Instrumental Considerations: Chandra or
XMM–Newton?
In this chapter I take a small step upwards in redshift and discuss the two
intervening WHIM filaments discovered by Nicastro et al. (2005a) through their
X-ray absorption lines in a high–quality Chandra spectrum of the blazar Mrk 421.
Although each of these two absorption systems was detected with high confidence
through multiple lines, the individual absorption lines were generally quite weakly
detected, mostly at the 2 − 4σ level. Moreover, while they employed high–quality
Chandra and FUSE data taken during exceptionally bright outbursts of Mrk 421,
the many archived XMM–Newton observations of this source were not included in
the analysis. With roughly twice the effective area of Chandra/LETG, XMM/RGS
is in principle superior for X-ray grating spectroscopy between ∼ 10 − 40 A;
however, its slightly worse resolution (approximately 60 mA FWHM, versus 50 mA
for Chandra/LETG), higher susceptibility to background flares, and multitude of
narrow instrumental features can present serious complications for WHIM searches.
114
Page 130
Independent confirmation of the Chandra results with a separate instrument
like XMM is thus important. While some groups have searched for WHIM features
in a limited number of XMM Mrk 421 spectra (e.g., Ravasio et al. 2005), a complete
and systematic analysis has yet to be performed. Here I present a search for z > 0
WHIM features employing all “good” observations of Mrk 421 available in the XMM
archive, and a comparison of these results to those presented by Nicastro et al.
(2005a).
5.1. Data Reduction and Measurements
We searched the XMM archive for all Mrk 421 Reflection Grating Spectrometer
(RGS) data. Although 31 separate observations were available, 16 had pointing
offsets ∆θ ∼> 60′′ while the rest were offset by less than 15′′. Since spectral resolution
and calibration quality can degrade at large offsets, we only included those with
∆θ < 15′′. One extremely short observation (0158971101, with texp = 237 s) was
also excluded to simplify the data reduction process. Using the standard XMM
Science Analysis System version 6.5.0 routines1, RGS1 light curves were built for the
remaining 14 “good” observations (see Table 5.1), and the spectra were reprocessed
to exclude periods of high background levels and coadded. These combined, filtered
RGS1 and RGS2 spectra have effective exposure times of ∼ 440 ks and over 9 × 106
combined RGS1+RGS2 first–order counts between 10 − 36 A with ∼ 15000 counts
1See xmm.vilspa.esa.es/sas
115
Page 131
per 0.06 A resolution element in RGS1 near 21 A, over twice that in the Nicastro et
al. (2005a) Mrk 421 Chandra spectrum.
We first used the spectral fitting program Sherpa2 to fit a simple power law
plus Galactic foreground absorption model to the RGS1 and RGS2 data; however,
at such high spectral quality the RGS response model does not appear to be
well–determined, and large residuals remained. For line measurements, we thus
only considered ∼ 2 A windows around each wavelength of interest, using a power
law to independently model the RGS1 and RGS2 continua within each window
and excluding the strongest narrow detector features (with typical widths of 70 mA
or less). None of the intervening absorption lines were apparent through a visual
inspection of the XMM spectrum, though several of the z = 0 lines reported by
Williams et al. (2005) could be seen clearly.
A narrow Gaussian absorption line (FWHM= 5 mA) was included in the
model for each line measurement or upper limit reported by Nicastro et al.
(2005a). When convolved with the RGS instrumental response these absorption
lines appeared broadened to the RGS line spread function (LSF) widths (typically
FWHM= 60− 70 mA; den Herder et al. 2001). The 2σ upper limits on all equivalent
widths were then calculated (allowing the central line wavelengths to vary within
the 1σ errors reported by Nicastro et al. 2005a). Since the shapes of the RGS1 and
RGS2 instrumental responses are quite different, these limits were calculated using
2cxc.harvard.edu/sherpa/
116
Page 132
both a joint fit to the RGS1+RGS2 spectra as well as the individual RGS1 and
RGS2 spectra. It should be noted that wherever one RGS unit is unusable, the
total response is effectively halved, at which point it has a similar effective area to
Chandra/LETG. The resulting equivalent width limits are listed in Table 5.2.
5.2. Discussion
Figure 5.1 shows the spectral windows used to determine upper limits on the
Nicastro et al. (2005a) measured lines, with the data (black), continuum fit (blue),
Chandra measurements and limits (Nicastro et al. 2005a, red solid and dotted lines
respectively), and XMM limits (green) overplotted. In all cases, the Nicastro et al.
(2005a) measurements (or 3σ upper limits) appear to be consistent with the 2σ
upper limits we have derived directly from the XMM data, as shown in the figure and
listed in Table 5.2. The Ovii line at z = 0.027 looks as though it might be visible
in the spectrum, but this is most likely due to the weak instrumental feature at
∼ 22.1 A. For two lines (Nvii and Nvi at z = 0.027) the XMM 2σ upper limits are
approximately equal to the Nicastro et al. (2005a) best–fit measurements, but since
the Nicastro et al. (2005a) values are quite uncertain this result is still consistent.
Why, then, with 2 − 4 times the counts per resolution element, was XMM
unable to detect the intervening absorption systems seen by Chandra? Several
factors appear to have been involved in this non–detection, primarily: (1) narrow
117
Page 133
instrumental features caused by bad detector columns, (2) the broader LSF as
compared to Chandra/LETG, and (3) fixed–pattern noise at long wavelengths:
1. While broad instrumental features can be taken into account by modifications
to the continuum model (as in Nicastro et al. 2005a), it is difficult or impossible
to distinguish narrow features from true absorption lines; thus, any line falling
near one of the detector features shown in Figure 5.2 can easily be lost3. This
was responsible for the non-detection of the z = 0.011 Ovii Kα line. Although
it was the strongest line reported by Nicastro et al. (2005a), its wavelength falls
directly on a narrow RGS1 feature and within the non–functional CCD4 region
on RGS2, thereby preventing this line from being detectable with either RGS.
Since 18% of the wavelength space for studying redshifted Ovii (λ > 21.6 A)
toward Mrk 421 is directly blocked by these narrow features (with this number
climbing to about 60% if resolution elements immediately adjacent to bad
columns are included), these bad columns present the single greatest hindrance
to XMM/RGS studies of the WHIM.
2. Even for lines where both RGS1 and RGS2 data are available and the
instrumental response appears to be relatively smooth, the lower resolution of
XMM contributes to the nondetectability of the weaker z > 0 absorption lines.
Figure 5.3 shows the LSFs for both XMM/RGS1 (solid) and Chandra/LETG
assuming an unresolved line with Wλ = 10 mA at 21.6 A. While the core of
3These response file data can be found at www.astronomy.ohio-state.edu/∼smita/xmmrsp/
118
Page 134
the RGS1 response is ∼ 20% broader than that of the LETG, the RGS1 LSF
has extremely broad wings: only 68% of the line flux is contained within the
central 0.1A of the RGS1 LSF, as opposed to 96% for the LETG. This reduces
the apparent depth of absorption lines by about a factor of two as compared
to Chandra/LETG, severely decreasing the line detectability.
3. At long wavelengths (λ ∼> 29 A) strong fixed–pattern noise is apparent as a
sawtooth pattern in the instrumental response, strongly impeding the detection
of species such as Nvi and Cvi. Indeed, in these wavelength regimes (the
lower two panels of Figure 5.1), the Nvi and Cvi absorption lines are nearly
indistinguishable from the continuum.
5.3. Disputed Results
After its publication, this paper understandably drew some criticism from
groups affiliated with the XMM mission in general, and the RGS instrument in
particular. Kaastra et al. (2006) and Rasmussen et al. (2006, hereafter R06), two
journal–submitted papers that subsequently appeared on the astro–ph preprint
server, contested our result that the XMM lacked sufficient sensitivity to detect the
intervening WHIM filaments reported by Nicastro et al. (2005a). In fact, both claim
that the lack of an XMM detection implies that the filaments are either a statistical
fluctuation or a result of instrumental artifacts in Chandra/LETG, and that X-ray
119
Page 135
absorption lines from the local WHIM have yet to be found. As Kaastra et al. (2006)
primarily focuses on a reanalysis of Chandra and XMM data while R06 directly
attempts to refute the points presented in this paper, I will primarily address the
R06 points here.
The primary claim of R06 is that custom reduction techniques, applied during
the processing of XMM/RGS data, allow narrow detector features to be modeled
to a high degree of accuracy. This would be quite an important development since,
as we described previously, such detector features prevent absorption lines at those
wavelengths from being accurately measured (or even detected in many cases). The
ability to successfully mitigate these features would open up substantially more of
the RGS wavelength range for the detection of redshifted WHIM lines.
However, perhaps the most surprising aspect of R06 is that, assuming the R06
fit is perfect, the upper limits measured for the two intervening OVII absorbers are, at
worst, in no more that weak conflict with the Nicastro et al. (2005a) measurements.
R06 presents column density upper limits calculated at the wavelengths of the two
redshifted Ovii absorption lines reported by Nicastro et al. (2005a), obtaining 90%
upper limits of NOVII < 1 and 1.1 × 1015 cm−2 for the z = 0.011 and z = 0.027
systems respectively. More stringent upper limits (< 6 × 1014 cm−2) are also
calculated when the wavelength is frozen to the value reported by Nicastro et al.
(2005a), but the error in the line position must be taken into account in order to
accurately calculate the equivalent width error–hence, the less stringent limits are
120
Page 136
correct. Nicastro et al. (2005a) report measurements of (1.0 ± 0.5) × 1015 cm−2 and
(0.7 ± 0.5) × 1015 cm−2 (also 90% errorbars) for these two systems. Clearly, the R06
upper limits are consistent with the R05 measurements. It is thus apparent that R06
reached their conclusion using the more stringent, but incorrectly–calculated, upper
limits.
It should also be noted that we have presented three serious problems with
using XMM/RGS for weak absorption line studies: not only the narrow detector
features, but also the broad LSF wings and the long–wavelength fixed–pattern noise.
R06 only attempts to refute one of these arguments, and then only over a very small
(1.3A) wavelength range–only 5% of the range considered in this paper. R06 does
not attempt to correct detector features near, or measure upper limits of, other
absorption lines reported by Nicastro et al. (2005a) such as Nvii, Nvi, and Cvi.
Until and unless techniques are developed to fully address all three of the factors
brought up here, the XMM/RGS will not be able to study the WHIM as effectively
as Chandra/LETG.
5.4. Conclusion
We have presented the highest signal–to–noise coadded XMM grating spectrum
of Mrk 421 to date, incorporating all available archival data. This spectrum serves
as an independent check on the recent detection of two z > 0 WHIM filaments
121
Page 137
by Nicastro et al. (2005a). While none of the Chandra–detected absorption lines
are seen in the XMM spectrum, the upper limits derived from the XMM data
are consistent with the equivalent widths reported by Nicastro et al. (2005a)
(even though the XMM data contain a larger number of counts), and hence do
not jeopardize the validity of the Chandra measurement. The non–detections can
be attributed primarily to narrow instrumental features in RGS1 and RGS2, as
well as the inferior spectral resolution of XMM and fixed–pattern noise at longer
wavelengths. This underscores the extreme difficulty of detecting the WHIM,
illustrates how the aforementioned (apparently small) effects can greatly affect the
delicate measurement of weak absorption lines, and re–emphasizes the importance of
high resolution and a smooth instrumental response function for current and future
WHIM absorption line studies.
122
Page 138
ID Date texpa tfilt
b Ratec
ks ks s−1
0099280101 2000 May 25 63.8 21.2 15.7
0099280201 2000 Nov 01 40.1 34.1 5.4
0099280301 2000 Nov 13 49.8 46.6 15.3
0099280501 2000 Nov 13 21.2 17.8 17.2
0136540101 2001 May 08 38.8 36.1 11.7
0136540301 2002 Nov 04 23.9 20.5 11.7
0136540401 2002 Nov 04 23.9 20.1 13.6
0136540701 2002 Nov 14 71.5 62.8 16.4
0136541001 2002 Dec 01 70.0 58.1 8.3
0158970101 2003 Jun 01 43.0 25.3 9.0
0158970201 2003 Jun 02 9.0 6.6 9.7
0158970701 2003 Jun 07 48.9 29.9 5.4
0158971201 2004 May 06 65.7 40.5 19.5
0162960101 2003 Dec 10 30.0 17.5 9.8
TOTAL 572.3 437.1 12.2
aTotal observation duration.bEffective RGS1 exposure time after filtering for
periods of high background levels.
cAverage count rate in the filtered RGS1 first–
order source spectral extraction region.
Table 5.1. XMM–Newton observation log
123
Page 139
Line λa za Wλ,N05aa Wλ,R1
b Wλ,R2b Wλ,R1+R2
b Note
A mA mA mA mA
Ne IXKα 13.80 ± 0.02 0.026 ± 0.001 < 1.5 < 5.2 < 1.9 < 2.9 1
O VIIKβ 19.11 ± 0.02 0.026 ± 0.001 < 1.8 < 2.5 < 2.1 < 1.5
O VIIIKα 19.18 ± 0.02 0.011 ± 0.001 < 4.1 < 7.6 < 5.8 < 4.1
O VIIIKα 19.48 ± 0.02 0.027 ± 0.001 < 1.8 · · · < 3.9 · · · 2
O VIIKα 21.85 ± 0.02 0.011 ± 0.001 3.0+0.9−0.8 · · · · · · · · · 2,3
O VIIKα 22.20 ± 0.02 0.028 ± 0.011 2.2 ± 0.8 < 3.9 · · · · · · 3
N VIIKα 25.04 ± 0.02 0.010 ± 0.001 1.8 ± 0.9 < 3.0 < 6.0 < 4.4
N VIIKα 25.44 ± 0.02 0.027 ± 0.001 3.4 ± 1.1 < 4.3 < 4.2 < 3.5
N VIKα 29.54 ± 0.02 0.026 ± 0.001 3.6 ± 1.2 < 3.8 < 8.7 < 3.4
C VIKα 34.69 ± 0.02 0.028 ± 0.001 2.4 ± 1.3 < 5.5 < 5.2 < 4.2
a Line wavelength, redshift, and equivalent width measurements (or 3σ upper limits) from
Nicastro et al. (2005a).
b2σ equivalent width upper limits measured from the RGS1 only (R1), RGS2 only (R2),
and joint (R1+R2) fits to the XMM–Newton spectrum, when available.
Note. — (1) A nearby chip gap in RGS1 renders this measurement unreliable, so only
the RGS2 measurement was used in Figure 5.1; (2) Line was unmeasurable in RGS1 because
of a detector feature; (3) Line was unmeasurable in RGS2 because of a detector feature.
Table 5.2. Absorption line equivalent width measurements
124
Page 140
Fig. 5.1.— Portions of the XMM–Newton RGS spectrum of Mrk 421 (black
histogram) with the continuum fit (blue), 2σ upper limit (green), and Nicastro et
al. (2005a) measurements and upper limits (red solid and dashed) overplotted. (0),
(1), and (2) denote z = 0, z = 0.011 and z = 0.027 lines respectively. Regions that
were excluded from the fit due to chip gaps and detector features are shaded; weaker
instrumental features are marked with vertical blue ticks.
125
Page 141
Fig. 5.2.— RGS1 (top panel) and RGS2 (bottom panel) instrumental response models
for the Ovii z = 0−0.5 region, as a function of wavelength (upper axes) and redshift
relative to λ = 21.602A (lower axes); a strong z = 0.25 line with NOVII = 1016 cm−2
(Wλ = 36 mA) is shown for reference.
126
Page 142
Fig. 5.3.— Comparison of the XMM RGS1 (solid) and Chandra HRC-S/LETG
(dotted) line spread functions for a Wλ = 10 mA unresolved absorption line at
21.602 A.
127
Page 143
Chapter 6
Summary and Future Work
6.1. Individual X-ray Sightlines
Three of the highest–quality Chandra grating spectra of extragalactic sources
have been analyzed in detail in an effort to determine the physical state of the z ≈ 0
warm–hot absorption lines tentatively observed in previous studies. Data from the
FUSE archive were also available for these objects, and these data were analyzed in
conjunction with the X-ray data. Two primary questions were addressed: (1) Is the
origin of the z = 0 X-ray absorption Galactic or extragalactic; and (2) Are the z = 0
X-ray lines associated with any of the Ovi low– or high–velocity components seen
along many quasar lines of sight?
The X-ray and UV data were folded through curve–of–growth and ionization
balance models to measure (or place limits on) the temperature, density, Doppler
parameter, and column density of the z = 0 gas along each line of sight. As it turns
out, none of the observed absorption systems can be conclusively placed at Galactic
or extragalactic distances from these data alone: although the lower limits on the
128
Page 144
absorber densities were consistent with an extragalactic medium (log ne > −4.7 for
Mrk 421, and > −5.5 for PKS 2155–304 and Mrk 279), a high–density, Galactic
medium could not be ruled out. For Mrk 421 and PKS 2155–304, where both
Oviii and Ovii could be measured, tight limits on the temperature of the absorber
(log T = 6.1 − 6.2 and log T = 6.18 − 6.36 respectively) were found from the
Oviii/Ovii ratio assuming pure collisional ionization.
The second question was addressed by directly comparing the inferred properties
of the X-ray absorbers with the fully resolved UV Ovi absorption. In the case
of Mrk 421, the Ovii absorption had a substantially lower velocity dispersion
than the low–velocity Ovi from the Galactic thick disk; the Ovi HVC, on the
other hand, was far too weak to coexist in the same gaseous phase as the Ovii
(requiring temperatures of log T > 7 to produce the observed Ovi(HVC)/Ovii
ratio). Likewise, the Ovii absorption toward Mrk 279 had a significantly higher
velocity dispersion than the low–velocity Ovi, and the Ovii centroid was about
2.5σ away from the Ovi HVC. In both of these cases, the Ovii does not appear to
arise in the same gaseous phase as any of the Ovi components. It is therefore likely
that the medium producing the Ovii absorption is either extragalactic or comprises
a previously uncharacterized Galactic component.
129
Page 145
6.2. The Importance of Spectral Fidelity
In Chapter 5, a direct comparison was made between long–duration observations
of Mrk 421 with both the XMM–Newton and Chandra grating spectrometers. Even
though this XMM/RGS spectrum had nearly four times the counts per resolution
element as the Chandra LETG spectrum (which, at the same nominal resolution,
should have resulted in roughly twice the sensitivity to WHIM lines), the two
redshifted WHIM absorption systems reported by Nicastro et al. (2005a) could
not be seen. Moreover, upper limits on the equivalent widths of these lines in
the XMM spectrum were consistent with the Chandra–measured strengths; in
short, despite having far better counting statistics, the XMM spectrum had worse
sensitivity than the Chandra spectrum. The difference is attributed to three primary
causes: a plethora of narrow detector features in XMM/RGS that render certain
wavelengths unobservable, broad wings on the RGS line–spread function that hamper
detectability, and fixed–pattern noise at longer wavelengths. This underscores
the need for a “clean” spectral response with a narrow line–spread function when
undertaking studies of weak absorption line systems such as the WHIM.
130
Page 146
6.3. Future Prospects
6.3.1. X-ray Observations
Although Chandra and FUSE have provided a great deal of insight into the
nature of the Galaxy’s (and/or Local Group’s) gaseous environment, they may
be nearing the end of what they can accomplish in this field. Only three or four
extragalactic X-ray sources (most of which were analyzed in this dissertation) are
bright enough to obtain an X-ray grating spectrum of reasonable quality in a few
hundred kiloseconds. As a result of Poisson statistics, four times the exposure time
are required to obtain a spectrum with twice the sensitivity; this additional time
adds up quickly, and time allocation committees are unwilling or unable to devote a
disproportionately large chunk of observing time to a single field. Even if such time
were awarded on Chandra, it is unlikely that the constraints on the local absorption
would be much better than those already achieved — with ∼ 700 km s−1 spectral
resolution at the Ovii wavelength, it is all but impossible to distinguish between
different velocity and temperature phases of the X-ray absorber, or to determine
which lower–ionization absorbers the Ovii is connected to.
There is certainly still plenty of room for interesting observations with Chandra,
of course: as mentioned in Chapter 4, better data on PKS 2155–304 would allow a
direct comparison of this sightline to Mrk 421. Medium–quality spectra of several
131
Page 147
more lines of sight would also be useful for determining the large–scale distribution
of the gas (though the amount of time required to obtain even one such quasar
spectrum is several hundred ksec). Thus, new X-ray observatories will be the key to
substantially moving this field forward observationally.
Two new missions, Constellation-X and XEUS, are on the horizon, but it is
yet unclear what their instrumental capabilities will ultimately be. To maximally
benefit WHIM science, it is critical that they have the following characteristics:
(1) Grating spectrographs, rather than calorimeters. Gratings have constant
wavelength resolution (hence higher velocity resolution at long wavelengths, where
redshifted WHIM systems are expected), while calorimeters have constant energy
resolution (thus losing velocity resolution at long wavelengths).
(2) High resolution (∆λ/λ of a few thousand). This dramatically increases the
prospects of detecting redshifted WHIM lines, and would also provide additional
information about the kinematics of the z ≈ 0 X-ray absorption.
(3) High quality spectral response. As shown in Chapter 5, in order for weak
absorption lines to be detectable, the detector must be free of narrow features that
can obliterate the real lines, and the line response function should be as narrow as
possible.
132
Page 148
6.3.2. Longer Wavelengths
Even in the absence of new X-ray data, other data have the potential to
contribute substantially to our understanding of the WHIM — particularly its
connection to low–redshift galaxies and the Galaxy itself. The WHIM does not
necessarily remain in a warm–hot state forever, and substantial quantities of it are
expected to cool and accrete onto galaxies. Such accretion may already be observed
in the form of Ovi and H i high–velocity clouds, though an unambiguous connection
between these and the WHIM has not yet been determined.
One potentially interesting set of objects is the population of compact
H i high–velocity clouds (CHVCs) seen more or less uniformly around the Galaxy.
These objects are typically unresolved with single dish radio telescopes, and their
distances and masses are unknown. Along with Smita Mathur, I have recently been
awarded time on the Spitzer Space Telescope to search for dust emission in three very
high column density CHVCs. Such a detection would imply that these objects are
most likely Galactic (since large quantities of dust emission require both relatively
high metallicity and proximity to the Galactic plane), while a non–detection with a
stringent upper limit would bolster the case for CHVCs being extragalactic.
133
Page 149
BIBLIOGRAPHY
Arav, N., Kaastra, J., Steenbrugge, K., Brinkman, B., Edelson, R., Korista, K. T.,
& de Kool, M. 2003, ApJ, 590, 174
Bennett, C. L., et al. 2003, ApJS, 148, 1
Cen, R., & Ostriker, J. P. 1999, ApJ, 514, 1
Chen, X., Weinberg, D. H., Katz, N., Dave, R. 2003, ApJ, 594, 42
Collins, J. A., Shull, J. M., & Giroux, M. L. 2005, ApJ, 623, 196
Courteau, S., & van den Bergh, S. 1999, AJ, 118, 337
Danforth, C. W., Howk, J. C., Fullerton, A. W., Blair, W. P., & Sembach, K. R.
2002, ApJS, 139, 81
Dave, R., et al. 2001, ApJ, 552, 473
den Herder, J. W., et al. 2001, A&A, 365, L7
Dickey, J. M., & Lockman, F. J. 1990, ARA&A, 28, 215
Drake, J., & Testa, P. 2005, Nature, 436, 525
Elvis, M., Wilkes, B. J., & Lockman, F. J. 1989, AJ, 97, 777
Elvis, M., Marengo, M., & Karovska, M. 2002, ApJ, 567, L107
Fang, T., Bryan, G. L., & Canizares, C. R. 2002, ApJ, 539, 532
Fang, T., Marshall, H. L., Lee, J. C., Davis, D. S., & Canizares, C. R. 2002, ApJ,
572, 127
Fang, T., Sembach, K. R., & Canizares, C. R. 2003, ApJ, 586, 49
134
Page 150
Ferland, G. J., 1996, Hazy, a Brief Introduction to Cloudy 90, University of
Kentucky, Physics Department Internal Report
Ferland, G. J., Korista, K. T., Verner, D. A., Ferguson, J. W., Kingdon, J. B., &
Verner, E. M. 1998, PASP, 110, 761
Garmire, G. P., Bautz, M. W., Ford, P. G., Nousek, J. A., Ricker, G. R. 2003, SPIE,
4851, 28
Grevesse, N., & Anders, E. 1989, AIP Conference Proceedings 183, 1
Haardt, F., & Madau, P. 1996, ApJ, 461, 20
Hellsten, U., Gnedin, N. Y., & Miralda–Escude, J. 1998, ApJ, 509, 56
Kaastra, J. S., Werner, N., den Herder, J. W. A., Paerels, F. B. S., de Plaa, J.,
Rasmussen, A. P., & de Vries, C. P. 2006, ApJ, submitted (astro-ph/0604519)
Kravtsov, A. V., Klypin, A., & Hoffman, Y. 2002, ApJ, 571, 563
Krongold, Y., Nicastro, F., Brickhouse, N. S., Elvis, M., Liedahl, D. A., & Mathur,
S. 2003, ApJ, 597, 832
Lennon, J. & McCartney, P. 1967, “With A Little Help From My Friends.” Lyrics,
Sgt. Pepper’s Lonely Hearts Club Band, Capitol Records.
Margon, B., Jones, T. W., & Wardle, J. F. C. 1978, AJ, 83, 1021
Marshall, H. L., Tennant, A., Grant, C. E., Hitchcock, A. P., O’Dell, S. L. &
Plucinsky, P. P. 2004, Proc. SPIE, 5165, 497
Mathur, S., Weinberg, D. H., & Chen, X. 2003, ApJ, 582, 82
McKernan, B., Yaqoob, T., Mushotzsky, R., George, I. M., & Turner, T. J. 2003,
ApJ, 589, L83
Miville–Deschenes, M.-A., Boulanger, F., Reach, W. T., & Noriega–Crespo, A. 2005,
ApJ, 631, L57
Murphy, E. M., Lockman, F. J., Laor, A., & Elvis, M. 1996, ApJS, 105, 369
Murray, S. S. & Chappell, J. H. 1985, SPIE, 597, 274
Nicastro, F., et al. 2002, ApJ, 573, 157
Nicastro, F., et al. 2003, Nature, 421, 719
135
Page 151
Nicastro, F., et al. 2005, ApJ, 629, 700
Nicastro, F., et al. 2005, Nature, 433, 495
Paerels, F., et al. 2001, ApJ, 546, 338
Penton, S. V., Stocke, J. T., & Shull, J. M. 2000, ApJS, 130, 121
Perna, R., & Loeb, A. 1998, ApJ, 503, L135
Pradhan, A. K. 2000, ApJ, 545, L165
Pradhan, A. K., Chen, G. X., Delahaye, F., Nahar, S. N., & Oelgoetz, J. 2003,
MNRAS, 341, 1268
Rasmussen, A. P., Kahn, S. M., Paerels, F., den Herder, J. W., Kaastra, J., & de
Vries, C. 2006, ApJ, submitted (astro-ph/0604515)
Ravasio, M., Tagliaferri, G., Pollock, A. M. T., Ghisellini, G., & Tavecchio, F. 2005,
A&A, 438, 481
Richter, P., Savage, B. D., Tripp, T. M., & Sembach, K. R. 2004, ApJS, 153, 165
Savage, B. D., et al. 2003, ApJS, 146, 125
Schmidt, M., Beiersdorfer, P., Chen, H., Thorn, D. B., Trabert, E., & Behar, E.
2004, ApJ, 604, 562
Sembach, K. R., et al. 2003, ApJS, 146, 165
Sembach, K. R., Tripp, T. M., Savage, B. D., & Richter, P. 2004, ApJS, 155, 351
Shull, J. M., Stocke, J. T., & Penton, S. 1996, AJ, 111, 72
Sternberg, A., McKee, C. F., & Wolfire, M. G. 2002, ApJS, 143, 419
Verner, D. A., Verner, E. M., & Ferland, G. J. 1996, Atomic Data Nucl. Data Tables,
64, 1
Wakker, B. P., et al. 2003, ApJS, 146, 1
Wang, Q. D., et al. 2005, ApJ, 635, 386
Weinberg, D. H., Miralda–Escude, J., Hernquist, L., & Katz, N. 1997, ApJ, 490, 564
Whittet, D.C.B. 1992, Dust in the Galactic Environment (Bristol: IOP)
136
Page 152
Williams, R. J., et al. 2005, ApJ, 631, 856
Williams, R. J., Mathur, S., & Nicastro, F. 2006, ApJ, 645, 179
Williams, R. J., Mathur, S., Nicastro, F., & Elvis, M. 2006, ApJ, 642, L95
137