THE NATURE OF THE [O III] EMISSION LINE SYSTEM IN THE BLACK HOLE
HOSTING GLOBULAR CLUSTER RZ2109
By
Matthew M. Steele

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Astrophysics & Astronomy
2012

ABSTRACT
THE NATURE OF THE [O III] EMISSION LINE SYSTEM IN THE BLACK
HOLE HOSTING GLOBULAR CLUSTER RZ2109
By
Matthew M. Steele
This work, focused on the description and understanding of the nature of a [O III]
emission line source associated with an accreting stellar mass black hole in a globlar cluster,
is comprised of three papers.
In the ﬁrst paper, we present a multi-facility study of the optical spectrum of the extragalactic globular cluster RZ2109, which hosts a bright black hole X-ray source. The optical
spectrum of RZ2109 shows strong and very broad [O III]λλ4959,5007 emission in addition
to the stellar absorption lines typical of a globular cluster. We use observations over an extended period of time to constrain the variability of these [O III] emission lines. We ﬁnd that
the equivalent width of the lines is similar in all of the datasets; the change in L[O III]λ5007
<
is ∼ 10% between the ﬁrst and last observations, which were separated by 467 days. The

velocity proﬁle of the line also shows no signiﬁcant variability over this interval. Using a simple geometric model we demonstrate that the observed [O III]λ5007 line velocity structure
can be described by a two component model with most of the ﬂux contributed by a bipolar
conical outﬂow of about 1,600 km s−1 , and the remainder from a Gaussian component with
a FWHM of several hundred km s−1 .
In the second paper, we present an analysis of the elemental composition of the emission line system associated with the black hole hosting globular cluster RZ2109 located in
NGC4472. From medium resolution GMOS optical spectroscopy we ﬁnd a [O III]λ5007/Hβ
emission line ratio of 106 for a 3200 km s−1 measurement aperture covering the full ve-

locity width of the [O III]λ5007 line, with a 95% conﬁdence level lower and upper limits
of [O III]λ5007/Hβ > 35.7 and < −110 (Hβ absorption). For a narrower 600 km s−1
aperture covering the highest luminosity velocity structure in the line complex, we ﬁnd
[O III]λ5007/Hβ = 62, with corresponding 95% conﬁdence lower and upper limits of > 30.2
and < −364. The measured [O III]λ5007/Hβ ratios are signiﬁcantly higher than can be
produced in radiative models of the emission line region with solar composition, and the
conﬁdence interval limits exclude all but the most extremely massive models. Therefore, we
conclude that the region from which the [O III]λ5007 emission originates must be hydrogen
depleted relative to solar composition gas. This ﬁnding is consistent with emission from an
accretion powered outﬂow driven by a hydrogen depleted donor star, such as a white dwarf,
being accreted onto a black hole.
In the third paper, we examine the variability of the [O III]λλ4959,5007 emission line
source in the NGC 4472 black hole hosting globular cluster RZ2109. Our continuing multifacility monitoring program ﬁnds the strong emission line source had decreased 24±2 percent
from the 2007–2010 mean levels in 2011 and 40±5 percent from the earlier mean in 2012. An
analysis of the variability of the emission line velocity proﬁle ﬁnds that the ﬂux ratio of higher
velocity 1600 km s−1 component to the lower velocity 300 km s−1 component has decreased
30 percent from 2009 to 2011, and the asymmetry between the red and blue wings of the
proﬁle has decreased 17 percent. We compare this variability to predictions of photoionized
nova ejecta models of the emission line region, and discuss its implications for an accretion
powered outﬂow from a CO WD-BH bianry model.

TABLE OF CONTENTS

List of Tables . . . . . . . . . . . . . . .
List of Figures

. . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . .

1 Introduction . . . . . . . . . . . . . . . . . .
1.1 BLACK HOLES . . . . . . . . . . . . .
1.1.1 X-Ray Binaries . . . . . . . . . .
1.1.2 Black Holes In Globular Clusters
1.2 RZ2109 . . . . . . . . . . . . . . . . . .
1.3 OUTLINE OF THIS STUDY . . . . . .

vii

. . . . . . . . . . . . . . . . . . viii
.
.
.
.
.
.

1
1
2
3
5
6

. . . . . . . . . . . . . . . . . . .

9

2 Velocity Structure and Variability of [O III] Emission in Black Hole Host
Globular Cluster RZ2109 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 OBSERVATIONS AND DATA REDUCTION . . . . . . . . . . . . . . . . .
2.2.1 Keck . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2 WHT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3 SOAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.4 Gemini . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.5 Data Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.6 Stellar Component Model . . . . . . . . . . . . . . . . . . . . . . . .
2.3 VARIABILITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 Equivalent Width Variability . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Line Proﬁle Variability . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 GEOMETRIC MODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5 DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.1 Variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.2 Structure of Emission Region . . . . . . . . . . . . . . . . . . . . . .
2.6 SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11
12
14
14
15
15
15
16
17
18
18
21
24
30
30
31
37

Bibliography . . . . . . . . . . . . . . .

iv

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

Bibliography . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

3 Composition of an Emission Line System in Black Hole
Cluster RZ2109 . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . .
3.2 OBSERVATIONS AND DATA REDUCTION . . . . . . .
3.3 ANALYSIS AND RESULTS . . . . . . . . . . . . . . . . .
3.3.1 Stellar Component Modeling . . . . . . . . . . . . .
3.3.2 Equivalent Widths . . . . . . . . . . . . . . . . . .
3.3.3 Emission Line System Modeling . . . . . . . . . . .
3.4 DISCUSSION AND CONCLUSION . . . . . . . . . . . . .
Bibliography . . . . . . . . . . . . . . .

40

Host Globular
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .

43
44
46
46
46
51
53
57

. . . . . . . . . . . . . . . . . . .

61

4 Variability of the [O III]λλ4959,5007 Emission
Host Globular Cluster RZ2109 . . . . . . . . .
4.1 INTRODUCTION . . . . . . . . . . . . . . .
4.2 OBSERVATIONS AND DATA REDUCTION
4.2.1 STIS . . . . . . . . . . . . . . . . . . .
4.2.2 Gemini . . . . . . . . . . . . . . . . . .
4.2.3 SOAR . . . . . . . . . . . . . . . . . .
4.2.4 Stellar Component Model . . . . . . .
4.3 VARIABILITY . . . . . . . . . . . . . . . . .
4.3.1 Equivalent Width Variability . . . . .
4.3.2 Velocity Proﬁle Variability . . . . . . .
4.3.3 VARIABILITY ANALYSIS . . . . . .
4.4 DISCUSSION . . . . . . . . . . . . . . . . . .

Line Source in Black Hole
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .

63
64
66
66
67
67
67
68
68
71
77
83

. . . . . . . . . . . . . . . . . . .

88

5 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Summary of Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Summary of Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

90
91
92

Bibliography . . . . . . . . . . . . . . .

95

Bibliography . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . .

A Emission Line Region Geometry and Velocity Proﬁles
A.1 Motivation and Code Description . . . . . . . . . . . . .
A.2 Spherical Outﬂow . . . . . . . . . . . . . . . . . . . . . .
A.3 Rotating Sphere . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

. 97
. 97
. 99
. 102

A.4
A.5
A.6
A.7
A.8
A.9

Disk-Like Outﬂow . . . . .
Rotating Disk . . . . . . .
Bipolar Conical Outﬂows .
Relativistic Eﬀects . . . .
Velocity Proﬁle Variability
General Observations . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

104
106
108
113
115
117

B Cloudy Modeling of [O III] Emission Line Regions
B.1 Description of Cloudy . . . . . . . . . . . . . . . . . .
B.2 Relevant Parameters . . . . . . . . . . . . . . . . . .
B.3 Solar Abundance Grid . . . . . . . . . . . . . . . . .
B.4 CO Models . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

118
118
119
124
126

LIST OF TABLES

Table 2.1

[O III]λ5007 Equivalent Widths . . . . . . . . . . . . . . . . . . . .

21

Table 3.1

Emission line equivalent widths and Hβ ratios . . . . . . . . . . . .

51

Table 3.2

[O III]λ5007/Hβ Ratio Limits . . . . . . . . . . . . . . . . . . . . .

56

Table 4.1

[O III]λ5007 Equivalent Widths . . . . . . . . . . . . . . . . . . . .

69

Table 4.2

Ripamonti & Mapelli (2012) Nova Model [O III]λ5007 Light Curves1

83

vii

LIST OF FIGURES

Figure 2.1

Figure 2.2

Figure 2.3

Figure 2.4

Figure 2.5

Gemini GMOS spectrum of black hole hosting globular cluster RZ2109.
The inset displays the [O III]λλ4959,5007 emission structure which
dominates the spectrum. The spectrum has been continuum ﬁt and
normalized, and smoothed with a three pixel boxcar function. . . . .

19

Gemini spectra internight variability. The lower (March 28, 2009),
middle (March 29, 2009), and upper (March 30, 2009) spectra have
been smoothed with a three pixel boxcar function and oﬀset on the
ﬂux axis for legibility. The dashed lines give the level of the normalized continuum for reference. . . . . . . . . . . . . . . . . . . . . . .

22

Keck and Gemini spectra comparison. The Gemini spectra (dotted
line) have been convolved with the instrumental resolution of the
Keck spectra (solid line.) The Keck observation preceded the Gemini
observation by 467 days.For interpretation of the references to color
in this and all other ﬁgures, the reader is referred to the electronic
version of this dissertation. . . . . . . . . . . . . . . . . . . . . . . .

23

Gemini spectra [O III] velocity structure. The top ﬁgure displays the
line proﬁle for [O III]λ5007, and the lower panel [O III]λ4959. The
two lines have been separated by a cut at an observed wavelength
of λ5006. As a result there may be a small degree of blending on
the extreme blue side of the [O III]λ5007 and the extreme red side of
[O III]λ4959. The dashed lines give the continuum level for reference.

25

Gemini spectra with geometric model. The solid black line shows
the Gemini spectra, the dotted line displays the best ﬁt bipolar conical outﬂow component of the geometric model, the dashed line the
Gaussian component of the model, and the solid gray line (red in the
electronic version) the full geometric model. . . . . . . . . . . . . . .

27

viii

Figure 2.6

Figure 3.1

Figure 3.2

Figure 3.3

Figure 3.4

Geometric model uncertainty contours. The top left plot displays the
uncertainty contours in opening and inclination angles with a ﬁxed
velocity of 1600 km s−1 . The top right plot give the uncertainty
contours in opening angle and velocity about a ﬁxed inclination of 65
deg. The bottom plot shows the uncertainty contours in velocity and
inclination angle about a ﬁxed opening angle of 70 deg. Contours in
each plot give the 2, 3, 4, and 5 sigma deviations from the Gemini
data. The plotted cross shows the position of the best ﬁt value. . . .

28

GMOS spectrum and the synthetic stellar component model. The
GMOS observations of RZ2109 are shown using the black line. The
gray line (red in the electronic version) depicts the best ﬁt synthetic
stellar model. The inset shows the wavelength region surrounding Hβ
in detail. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

47

Stellar Model Fitting Uncertainties and Hβ equivalent width contours. The gray scale grid displays the probability density in stellar
parameter space that the stellar component of the RZ2109 spectrum
would be determined to have a given set of parameters as described
in section 3.3.1. The contour lines show the measurement of the Hβ
equivalent width with a 600 km s−1 aperture for a given stellar component model relative to the best ﬁt model, as described in section
3.3.2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

49

Equivalent Width apertures. The upper panel displays the Hβ region
of the GMOS spectrum in velocity space, and the lower panel gives
the [O III]λ5007 region. The 600 km s−1 aperture over which the
equivalent width was measured is shaded with dark gray. The 3200
km s−1 aperture is shown in light gray. The dashed line gives the
continuum level. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

50

[O III]λ5007/Hβ ratio contours in Cloudy model parameter space
of a solar composition gas. The Cloudy models depend on three
free parameters: inner radius of the gas distribution (R0 ), hydrogen
density at the inner radius (ρH (R0 )), and total gas mass. Panel (a)
shows the [O III]λ5007/Hβ ratio contours for 0.2 M⊙ gas mass. At
this mass the maximum ratio of 26 occurs at position indicated by the
cross. A maximum ratio of [O III]λ5007/Hβ =34 for all parameter
space occurs for a gas mass of 100 M⊙ with contours for the other
two free parameters shown in panel (b). . . . . . . . . . . . . . . .

54

ix

Figure 3.5

Figure 4.1

Maximum [O III]λ5007/Hβ ratios by gas mass. The trend-line depicts
the maximum [O III]λ5007/Hβ ratio from Cloudy models for any
given total gas mass, allowing the other free parameters (inner radius
of the gas distribution and hydrogen density at the inner radius) to
vary as necessary. The maximum ratio occurs with a total gas mass
of ∼ 100 M⊙ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

56

RZ2109 [O III]λλ4959,5007 Emission Line Complex. The black line
displays the 2009 Gemini observation, the red line gives the 2011
Gemini observation. Both spectra were collected using the same instrumental setup and similar integration times. Note that the [O III]
ﬂux is lower at all wavelengths in the 2011 observation. . . . . . . .

70

Figure 4.2

Scaled [O III]λλ4959,5007 Emission Line Complex. The spectra displayed in this plot have been normalized to their respective [O III]λ5007
equivalent width measurements to allow a comparison of the relative
velocity proﬁles of the [O III]λλ4959,5007 complex. The black line
gives the 2009 Gemini observation, the red line displays the 2011
Gemini data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Figure 4.3

Statistical Signiﬁcance of [O III]λλ4959,5007 proﬁle variability. The
black line give the statistical signiﬁcance of the diﬀerence between the
2009 and 2011 Gemini observations of the RZ2109 [O III]λλ4959,5007
line proﬁle in 5 ˚ bins. For reference the 2009 observation is given
A
as a blue dotted line, the 2011 observation as a green dotted line. . .

74

Geometric Model Velocity Proﬁle. The Steele et al. (2011) geometric
line proﬁle model is ﬁt to the 2011 Gemini [O III]λ5007 velocity proﬁle
(black line). All geometric parameters are the same as found by
Steele et al. (2011) for the 2009 Gemini data. The bipolar conical
outﬂow (blue line) has an opening angle of 70 degrees, inclination of
65 degrees, and velocity of 1600 km s−1 . The receding/approaching
outﬂow ratio is 1.15. The Gaussian component (green line)has a
FWHM of 310 km s−1 . The two components are co-added to produce
the modeled velocity proﬁle displayed using a red line. . . . . . . . .

75

X-ray and [O III]λ5007 light curves. The x-ray ﬂux measurements
displayed in the top panel are taken from Maccarone et al. (2010).
The [O III]λ5007 equivalent width measurements are published in
Steele et al. (2011) and the present work. . . . . . . . . . . . . . . .

78

Figure 4.4

Figure 4.5

x

Figure 4.6

Velocity Proﬁle Variability. The bipolar conical outﬂow model line
proﬁle is shown under the inﬂuence of a variable photoionizing source.
The Maccarone et al. (2010) light curve has been approximated using a step function with high luminosity (4 × 1039 erg s−1 ) and low
luminosity (1 × 1038 erg s−1 ) regimes. The diﬀerent proﬁles result
from the diﬀerent light travel time between regions of the emission
line region and the observer. The solid red line indicates the initial
line proﬁle, the solid black line at time equal to 0.5 light crossing time
of the radius, the dashed line after 1.0 light crossing time, the dashed
and dotted line 1.5 light crossing times, and the dotted line 2.0 light
travel times. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

80

Figure A.1

Spherical Outﬂow. All models have a radially directed outﬂow velocity of 1000 km s−1 . In all panels the solid line indicates a inclination
of the symmetry axis with respect to the observer’s line of sight of
0 degrees, the dashed line 45 degrees, and the dotted line 90 degrees. Panel (a) displays an model with uniform emissivity, in panel
(b) the emissivity is proportional to the sine of the inclination above
the mid-plane, for panel (c) an external ionizing source produces an
emissivity function that decreases linerarly with distance from the
incident surface of the ionizing ﬂux. . . . . . . . . . . . . . . . . . . 100

Figure A.2

Rotating sphere. All models have velocities of 1000 km s−1 and radii
of 10 percent of the outer radius, except panel (d) where the velocity
is 1000 km s−1 at the outer radius. Solid lines indicate inclination of
the symmetry axis relative to the observers line of sight of angles of 0
degrees, dashed lines 45 degrees, and dotted lines 90 degrees. Figure
(a) α = 2 (inverse square emissivity) β = 0.5 (Keplerian rotation),
(b) α = 0 (constant emissivity) β = 0.5, (c) α = 2 β = 0 (constant
velocity), (d) α = 2 β = −1 (solid body rotation). . . . . . . . . . . 103

Figure A.3

Disk-like outﬂow. All models have a radially directed velocity of 1000
km s−1 . Solid lines indicate an inclination of 0 degrees, dashed lines
45 degrees, and dotted lines 90 degrees. The models in panel (a) have
a truncation angle of 2.5 degrees above the mid-plane, panel (b) has
a truncation angle of 10 degrees. . . . . . . . . . . . . . . . . . . . 105

Figure A.4

Rotating disk. All models have velocities of 1000 km s−1 and radii of
10 percent of the outer radius. Solid lines indicate inclination angles
of 0 degrees, dashed lines 45 degrees, and dotted lines 90 degrees.
Figure (a) α = 2 (inverse square emissivity) β = 0.5 (Keplerian
rotation), (b) α = 0 (constant emissivity) β = 0.5, (c) α = 2 β = 0
(constant velocity), (d) α = 2 β = −1 (solid body rotation). . . . . 107
xi

Figure A.5

Bipolar conical outﬂow. The velocity proﬁle of models with opening
angles of 10 degrees. Panel (a) displays an inclination angle of 0
degrees, panel (b) 30 degrees, panel (c) 60 degrees, and panel (d) 90
degrees. All models have outﬂow velocities of 1000 km s−1 . . . . . . 109

Figure A.6

Bipolar conical outﬂow. The velocity proﬁle of models with opening
angles of 40 degrees. All other model parameters are identical to
those in ﬁgure A.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

Figure A.7

Bipolar conical outﬂow. The velocity proﬁle of models with opening
angles of 80 degrees. All other model parameters are identical to
those in ﬁgure A.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Figure A.8

Bipolar conical outﬂow. The velocity proﬁle of models with opening
angles of 120 degrees.All other model parameters are identical to those
in ﬁgure A.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

Figure A.9

Relativistic eﬀects on a bipolar conical outﬂow. All velocity proﬁles
represent a model with an opening angle of 35 degrees, an inclination
of 35 degrees and and inclination of 65 degrees. The solid black line
has a outﬂow velocity of 1 percent the speed of light, the dashed line
10 percent and the dotted line 20 percent. The velocity axis has been
normalized to the outﬂow velocity of each model. . . . . . . . . . . 114

Figure A.10 Variability of bipolar conical outﬂow. The velocity proﬁles of lines
displaying a one magnitude drop in emissivity, representing a corresponding decrease in the incident ionizing ﬂux. All models have an
opening angle of 35 degrees and a outﬂow velocity of 1000 km s−1 .
In each panel the solid red line indicates the initial line proﬁle, the
solid black line at time equal to 0.5 light crossing time of the radius,
the dashed line after 1.0 light crossing time, the dashed and dotted
line 1.5 light crossing times, and the dotted line 2.0 light travel times.
Panel (a) depicts an inclination of 0, (b) 30 degrees, (c) 60 degrees,
and (d) 90 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Figure B.1

Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.1–31.2 . . . . . 128

Figure B.2

Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.3–31.4 . . . . . 129

Figure B.3

Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.5–31.6 . . . . . 130

Figure B.4

Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.7–31.8 . . . . . 131
xii

Figure B.5

Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.9–32.0 . . . . . 132

Figure B.6

Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.1–32.2 . . . . . 133

Figure B.7

Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.3–32.4 . . . . . 134

Figure B.8

Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.5–32.6 . . . . . 135

Figure B.9

Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.7–32.8 . . . . . 136

Figure B.10 Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.9–33.0 . . . . . 137
Figure B.11 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.1–33.2 . . . . . 138
Figure B.12 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.3–33.4 . . . . . 139
Figure B.13 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.5–33.6 . . . . . 140
Figure B.14 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.7–33.8 . . . . . 141
Figure B.15 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.9–34.0 . . . . . 142
Figure B.16 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.1–34.2 . . . . . 143
Figure B.17 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.3–34.4 . . . . . 144
Figure B.18 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.5–34.6 . . . . . 145
Figure B.19 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.7–34.8 . . . . . 146
Figure B.20 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.9–35.0 . . . . . 147
Figure B.21 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.1–35.2 . . . . . 148
Figure B.22 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.3–35.4 . . . . . 149
Figure B.23 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.5–35.6 . . . . . 150
Figure B.24 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.7 . . . . . . . . 151
xiii

Figure B.25 Solar Abundance Grid Emissivity. log M = 32.5–32.8 . . . . . . . . 152
Figure B.26 Solar Abundance Grid Emissivity. log M = 32.9–33.2 . . . . . . . . 153
Figure B.27 Solar Abundance Grid Emissivity. log M = 33.3–33.6 . . . . . . . . 154
Figure B.28 Solar Abundance Grid Emissivity. log M = 33.7–34.0 . . . . . . . . 155
Figure B.29 Solar Abundance Grid Emissivity. log M = 34.1–34.4 . . . . . . . . 156
Figure B.30 Solar Abundance Grid Emissivity. log M = 34.5–34.8 . . . . . . . . 157
Figure B.31 Solar Abundance Grid Emissivity. log M = 34.9–35.2 . . . . . . . . 158
Figure B.32 Solar Abundance Grid Emissivity. log M = 35.3–35.6 . . . . . . . . 159
Figure B.33 Solar Abundance Grid Emissivity. log M = 35.7 . . . . . . . . . . . 160
Figure B.34 CO Enriched Models. . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Figure B.35 CO Enriched Models. . . . . . . . . . . . . . . . . . . . . . . . . . . 162

xiv

Chapter 1
Introduction
The globular cluster RZ2109 plays host to a combination of member components which are,
as of this writing, without known peer. The cluster itself is located in the elliptical galaxy
NGC 4472 situated at a distance of 16 Mpc in the Virgo Galaxy Cluster. What makes
RZ2109 unique is its coincidence with a luminous x-ray source, and a pair of [O III] emission
lines that appear in its optical spectrum. It is the origin of those two emission lines, and
more speciﬁcally the nature of the system from which they emanate that is the subject of
this work.

1.1

BLACK HOLES

Black holes are a near ubiquitous feature of astrophysical systems from the stellar scale to the
galactic. From solar mass black holes of a few solar masses to supermassive black holes of 109
solar masses, few other physical phenomenon cover the same broad range in scale. Perhaps
even more staggering is the expectation that black holes all the way along the mass spectrum
behave in similar ways, following simple scaling relations for range of properties. As observers
1

and theoreticians of stellar mass black holes are fond of pointing out, dynamic evolution of
black hole accretion systems that occur on timescales an appreciable fraction of the age of the
universe in supermassive systems have timescales of hours to days in stellar mass systems.
Similarly one can play a numbers game when selecting a mass range of black holes to study.
Galaxies host a single supermassive black hole, and by even the most conservative estimates
orders of magnitude more stellar mass systems. Stretching the argument further theoretical
predictions touched upon in following sections predict that black hole x-ray binaries are
preferentially produced in dense stellar environments like globular clusters. Furthermore in
comparison to other local group galaxies, the Milky Way has a rather small globular cluster
population. Therefore, it has been modestly argued, a promising avenue for the advancement
of understanding of black holes is the study of extragalactic black holes. The globular cluster
RZ2109 is one such object.

1.1.1

X-Ray Binaries

Binary stars are comprised of two stars orbiting around their shared center of gravity. In
an x-ray binary one of the pair is a compact object; either a neutron star or black hole.
X-ray binaries occur in multiple forms and classiﬁcations usually grouped by the mass of the
companion to the compact object. The two major subgroups of x-ray binaries are low mass
x-ray binaries (LMXB) and high mass x-ray binaries (HMXB). As suggested by the naming
schemes, the companion stars in LMXB are stars with low mass (M

2M⊙ ) and Be or

supergiant stars in HMXB. Both categories of x-ray binaries are most frequently identiﬁed
via the x-ray emission produced through their respective accretion processes.
In LMXB, the category of x-ray binary most relevant for this work, the accretion is
2

produced when the gravitational pull from the compact object strips the outer layers of the
companion star’s atmosphere causing it to overﬂow the volume deﬁned by the companions
gravitational binding, known as the Roche lobe. The gas then forms an accretion disk around
the black hole where the viscous dissipation of the gravitational energy heats the gas to a
temperature at which it is an x-ray source. The x-ray luminosities from these LMXB range
from 1033 –1040 erg s−1 , and can vary on timescales from hours up to a decade (McClintock &
Remillard, 2006). A sub class of these objects are know to drive radio synchrotron outﬂows,
a few which produce jets at velocities of tens of percent the speed of light (Mirabel &
Rodr´
ıguez, 1999).
While there may be other means of detecting stellar mass black holes, such as gravitational lensing of background sources by isolated black holes (Bennett et al., 2002), to date
no unambiguous detections of stellar mass black holes have been made apart from those in
x-ray binaries.

1.1.2

Black Holes In Globular Clusters

Globular clusters are gravitationally bound collections of 104 –107 stars. These clusters provide some of the most dense stellar environments known and have been used throughout
the last century alternatively as pristine environments for studying early stellar populations
and exotic environments which produce novel astrophysical systems. The Milky Way galaxy
has around 160 known globular clusters arranged in a roughly spherical distribution. This
cluster population is an order of magnitude smaller than in other galaxies in the local group,
so numerous studies interested in cluster demographics and unique cluster member objects
turn to extragalactic cluster systems as targets of study.
3

For much of the past 30 years theoretical work on black holes in globular clusters has
centered around the study of cluster stellar dynamics. Based on both theoretical (Spitzer
& Mathieu, 1980) and observational (Gunn & Griﬃn, 1979) considerations globular clusters
were initially thought to be deﬁcient in binaries compared to Galactic ﬁeld populations, either due to diﬀerences in the initial binary formation rate or their subsequent destruction
by dynamic processes. Even once the frequency of binaries and x-ray binaries and their
importance in cluster dynamics was established (Hut et al., 1992, and references therein),
considerable doubt remained as to whether a black hole x-ray binary could exist in a dynamically stable state in a globular cluster. Sigurdsson & Hernquist (1993) and Kulkarni
et al. (1993) suggested that dynamical interactions in the cores of globular clusters would
eject black holes from the system on the order of the dynamical time scale. Contrary to
these arguments more recent work has suggested that dynamics of dense cluster cores open
channels for black hole binary formation unavailable in galactic ﬁelds, such that clusters may
contain a higher occurrence of these x-ray sources than other environments (Ivanova et al. ,
2010).

It is only within the past ﬁve years that attention has turned from questions of whether
dynamically stable of black hole systems could exist in globular clusters to questions pertaining to the identiﬁcation, demographics, and environmental interactions of black hole hosting
globular clusters. Maccarone et al. (2007) provided the ﬁrst unambiguous identiﬁcation of a
black hole in a globular cluster, citing a highly luminous x-ray source with variability suﬃcient to rule out an ensemble of less luminous neutron star binaries as a source candidate.
This basic approach of identifying a high luminosity x-ray source with strong variability has
subsequently identiﬁed a growing number of black holes in globular clusters (Brassington
4

et al., 2010; Shih et al., 2010; Maccarone et al., 2011). One additional candidate has been
identiﬁed on the basis of the x-ray luminosity and spectrum (Irwin et al., 2010), thought the
classiﬁcation of this object remains unsettled.

1.2

RZ2109

The Maccarone et al. (2007) black hole is located in the globular cluster RZ2109 as designated by the Rhode & Zepf (2001) catalog of the globular cluster system of NGC 4472.
Follow up spectroscopic observations by Zepf et al. (2007) of the x-ray source’s optical counterpart to conﬁrm its globular cluster identiﬁcation revealed the presence of an emission line
corresponding with [O III]λ5007 . Emission lines sources in globular clusters, while rare,
are not unknown. A very small fraction of Galactic and extragalactic globular clusters host
planetary nebulae (Jacoby et al., 1997; Larsen, 2008) or supernovae remnants (Chomiuk et
al., 2008) which produce [O III] emission lines.
Higher resolution spectra published by Zepf et al. (2008) eliminated the possibility of
either of the classes of objects producing the observed emission. The RZ2109 source displayed
a broad [O III]λ5007 line a width of at least 3000 km s−1 , compared to the tens of km s−1 for
planetary nebulae. Beyond the [O III]λλ4959,5007 emission complex the Zepf et al. (2008)
spectrum displayed no additional emission lines, whereas supernovae remnants would display
several addition strong emission lines produced by heavier elements such as Ne. Zepf et al.
(2008) argued that the [O III] emission lines likely originated from a wind driven across the
cluster and photoionized by the x-ray source.
But the nature of this wind, its structure, composition, or even its connection to the
RZ2109 black hole were unclear. Furthermore since the RZ2109 black hole was the ﬁrst
5

unambiguous identiﬁcation of a stellar mass black hole in a globular cluster, the possibility
that other black hole hosting clusters might also harbor similar emission line regions seems
plausible. If true these environments could potentially provide an promising laboratory
for the study of an accreting black hole’s interactions with a dense stellar environment.
Generating an understanding of the properties of the RZ2109 [O III] emission line system is
a promising ﬁrst step in developing the tools and techniques for identifying and analyzing
this potential class of objects.

1.3

OUTLINE OF THIS STUDY

The body of this work occurs in three chapters each of which takes the form of a paper. The
ﬁrst paper—contained in chapter two—focuses on the RZ2109 [O III]λλ4959,5007 emission
line strength, velocity proﬁle, and variability for a series of observations from 2007–2009 and
constructs a simple model for the structure of the emission line region. In chapter three
the second paper addresses limits on the Hβ emission line and performs synthetic spectral
modeling of the emission line region to constrain its gas composition. Chapter four, the third
paper, updates the [O III]λλ4959,5007 emission line strength and variability analysis with
observations from 2009–2012 and uses the detected variability to constrain proposed models
for the emission line region.
Following the body of the text, two appendices are included. The ﬁrst of these documents
the emission line velocity proﬁle modeling code developed to describe the geometry of the
RZ2109 emission line region. In the second the calculations preformed to generate the
synthetic spectrum used to constrain the composition of the RZ2109 emission line region are
detailed, along with a brief guide to the use of the Cloudy code (Ferland et al., 1998) for
6

this purpose.

7

BIBLIOGRAPHY

8

BIBLIOGRAPHY

Bennett, D. P., Becker, A. C., Quinn, J. L., et al. 2002, ApJ, 579, 639
Brassington, N. J., et al. 2010, arXiv:1003.3236
Chomiuk, L., Strader, J., & Brodie, J. P. 2008, AJ, 136, 234
Ferland, G. J., Korista, K. T., Verner, D. A., et al. 1998, PASP, 110, 761
Gunn, J. E., & Griﬃn, R. F. 1979, AJ, 84, 752
Hut, P., McMillan, S., Goodman, J., et al. 1992, PASP, 104, 981
Kulkarni, S. R., Hut, P., & McMillan, S. 1993, Nature, 364, 421
Irwin, J. A., Brink, T. G., Bregman, J. N., & Roberts, T. P. 2010, ApJ, 712, L1
Ivanova, N., Chaichenets, S., Fregeau, J., et al. 2010, ApJ, 717, 948
Jacoby, G. H., Morse, J. A., Fullton, L. K., Kwitter, K. B., & Henry, R. B. C. 1997, AJ,
114, 2611
Larsen, S. S. 2008, A&A, 477, L17
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2007, Nature, 445, 183
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2011, MNRAS, 410, 1655
9

McClintock, J. E., & Remillard, R. A. 2006, Compact stellar X-ray sources, 157
Mirabel, I. F., & Rodr´
ıguez, L. F. 1999, ARA&A, 37, 409
Rhode, K. L., & Zepf, S. E. 2001, AJ, 121, 210
Shih, I. C., Kundu, A., Maccarone, T. J., Zepf, S. E., & Joseph, T. D. 2010, ApJ, 721, 323
Sigurdsson, S., & Hernquist, L. 1993, Nature, 364, 423
Spitzer, L., Jr., & Mathieu, R. D. 1980, ApJ, 241, 618
Zepf, S. E., Maccarone, T. J., Bergond, G., Kundu, A., Rhode, K. L., & Salzer, J. J. 2007,
ApJ, 669, L69
Zepf, S. E., et al. 2008, ApJ, 683, L139

10

Chapter 2
Velocity Structure and Variability of
[O III] Emission in Black Hole Host
Globular Cluster RZ2109
Matthew M. Steele 1 , Stephen E. Zepf 1 ,Arunav Kundu 1 2 , Thomas J. Maccarone3 ,
Katherine L. Rhode 4 , and John J. Salzer 4

We present a multi-facility study of the optical spectrum of the extragalactic globular
cluster RZ2109, which hosts a bright black hole X-ray source. The optical spectrum of
RZ2109 shows strong and very broad [O III]λλ4959,5007 emission in addition to the stellar
1 Department of Physics & Astronomy, Michigan State University, East Lansing, MI

48824; e-mail: steele24@msu.edu
2 Eureka Scientiﬁc, 2452 Delmer Street Suite 100, Oakland, CA 94602-3017
3 School of Physics & Astronomy, University of Southampton, Southampton, Hampshire
S017 1BJ, UK
4 Department of Astronomy, Indiana University, Bloomington, IN 47405
This chapter was originaly published as: Steele, M. M., Zepf, S. E., Kundu, A., et al.
2011, ApJ, 739, 95
11

absorption lines typical of a globular cluster. We use observations over an extended period of
time to constrain the variability of these [O III] emission lines. We ﬁnd that the equivalent
<
width of the lines is similar in all of the datasets; the change in L[O III]λ5007 is ∼ 10%

between the ﬁrst and last observations, which were separated by 467 days. The velocity
proﬁle of the line also shows no signiﬁcant variability over this interval. Using a simple
geometric model we demonstrate that the observed [O III]λ5007 line velocity structure can
be described by a two component model with most of the ﬂux contributed by a bipolar
conical outﬂow of about 1,600 km s−1 , and the remainder from a Gaussian component with
a FWHM of several hundred km s−1 .

KEYWORDS: galaxies: individual (NGC 4472) — galaxies: star clusters: individual
(NGC 4472) — globular clusters: general — X-rays: binaries — X-rays: galaxies: clusters

2.1

INTRODUCTION

The ﬁrst unambiguous black hole X-ray source in a globular cluster was discovered by Maccarone et al. (2007) in the extragalactic globular cluster RZ2109. This globular cluster is
in the Virgo elliptical galaxy NGC 4472 (M49), and is a luminous, metal-poor cluster, with
an absolute magnitude of MV = −10.0 and a B-V color of 0.68, located 6.6′ away from
the center of its host galaxy (Rhode & Zepf, 2001). The black hole X-ray source in RZ2109
was discovered in XMM observations of the NGC 4472 globular cluster system, of which
RZ2109 is a member. The XMM observations of the source XMMU122939.7+075333 in
RZ2109 showed that it had a X-ray luminosity of ≃ 4 × 1039 ergs s−1 with x-ray counts that
varied by a factor of 7 over a few hours (Maccarone et al., 2007). The high X-ray luminosity,
12

more than an order of magnitude higher than the Eddington luminosity for a neutron star,
requires either a black hole or multiple neutron stars in the old population of the globular
cluster. The strong variability within the XMM observation rules out multiple neutron stars,
and indicates that the source is a black-hole system (Maccarone et al., 2007).
A study of existing optical spectra of RZ2109 revealed that it had broad, strong [O III]λ5007
emission at the absorption line radial velocity of the globular cluster (Zepf et al., 2007).
Follow-up optical spectroscopy with the Keck telescope showed stellar absorption lines typical of an old globular cluster and remarkably broad and luminous [O III]λλ4959,5007 emission
lines (Zepf et al., 2008, hereafter Z08). Speciﬁcally, Z08 showed that the velocity width of the
[O III]λ5007 line is ≃ 2, 000 km s−1 , and that the line luminosity is about 1.4×1037 ergs s−1 .
Moreover, there is no sign of emission lines other than [O III], and the L([O III])/L(Hβ) ratio
appears to be at least 30 (Z08).
The observed high luminosity and broad velocity width of the [O III]λλ4959,5007 emission
lines place strong constraints on the origin of these emission lines and the nature of the
black hole X-ray source in RZ2109. For example, the presence of this very strong and broad
[O III] emission in the black hole X-source hosting globular cluster RZ2109 and the absence
of similar emission in any other spectroscopic study of globular clusters argues strongly
against signiﬁcant beaming of the X-rays (Gnedin et al., 2009). Furthermore, Z08 showed
that the observed broad [O III]λλ4959,5007 velocities cannot be explained by gravitational
motions near the black hole because the available volume suﬃciently close to the black hole is
many orders of magnitude too small to produce the observed [O III]λ5007, given the critical
density of the line. Instead, the broad velocity and high luminosity require a strong outﬂow
from the accreting black hole. Z08 suggests that such strong outﬂows occur for systems near
13

their Eddington limit (see also Proga, 2007, and references therein). The observed LX of
≃ 4 × 1039 ergs s−1 thus indicates a stellar mass for the accreting black hole in the globular
cluster RZ2109 (see discussion in Z08).
In this study we present several sets of new optical spectroscopic data on the broad
[O III]λλ4959,5007 lines in RZ2109. We use the multiple observations at diﬀerent times to
study the variability of the [O III]λλ4959,5007 emission over baselines ranging from one to
467 days. We also take advantage of the higher signal-to-noise and higher spectral resolution
of our new Gemini data to constrain models of the [O III]λλ4959,5007 emitting regions.

2.2

OBSERVATIONS AND DATA REDUCTION

In this study we use medium resolution optical spectra of RZ2109 from four observatories.
One set of observations has been previously presented, and the other three are described for
the ﬁrst time here. The four sets of observations span an interval of 467 days from the initial
to the most recent observation.

2.2.1

Keck

Optical spectra were obtained using the Low Resolution Imaging Spectrograph (Oke et al.,
1995) on the Keck Telescope. These data, originally reported in Z08, have a wavelength
coverage of 3200–5500, 5800–8930 ˚ and a measured spectral resolution of R ∼ 400. The
A
spectra were collected on the nights of UT 2007 December 17–18.
14

2.2.2

WHT

Observations were made on UT 2008 January 5–6 using the 4.2 m William Herschel Telescope
(WHT) with the Intermediate dispersion Spectrograph and Imaging System operating in
longslit mode. The WHT spectra have a wavelength coverage of 3900–5350, 5390–10000 ˚
A
and a measured resolution of R ∼ 660. To produce these observations the spectrograph was
set up using the 1.53 arcsec slit with the R300B grating in the blue arm and the R158R
grating in the red arm. For purposes of calibration the standard star HZ44 was observed
using an identical instrumental setup.

2.2.3

SOAR

Observations from Goodman High Throughput Spectrograph (Clemens et al., 2004) on the
4.1m Southern Astrophysical Research Telescope (SOAR) were collected on UT 2009 January
23, 24 and February 22. The SOAR spectra have a wavelength coverage of 4352–7024 ˚ and
A
a measured resolution of R ∼ 1600, using the KOSI 600 grating and a slit width of 0.84
arcsec.

2.2.4

Gemini

Spectra from the Gemini South Telescope were obtained under a queue observation using
the Gemini Multi-Object Spectrograph (Hook et al., 2004) in longslit mode. The data were
obtained on the nights of UT 2009 March 28, 29 and 30 (program GS-2009A-Q-1). The
instrumental setup included the use of the B1200 grating with a slit width of 1.0 arcsec.
This setup produced a wavelength coverage for the resulting spectra of 4445–6023 ˚ with a
A
measured spectral resolution of R ∼ 2400.
15

2.2.5

Data Reduction

All data presented for the ﬁrst time in this work were reduced using standard IRAF NOAO
longslit tools, including the Gemini IRAF package for the Gemini data. The processed two
dimensional spectra were extracted and, except where noted in Section 2.3.2, the data from
the same facility obtained on consecutive nights were co-added to increase the signal-tonoise ratio of each observation. The low signal-to-noise of the SOAR observations due to the
smaller instrument aperture and relative ineﬃciency of the spectrograph at the time necessitated co-addition of the two dimensional spectra prior to extraction of the one dimensional
spectrum. The WHT spectrum was ﬂux calibrated using observations of standard star HZ44;
all other spectra were left uncalibrated and simply normalized to a polynomial continuum
ﬁt. For the [O III] line proﬁle measurements and modeling presented in this work, simple
continuum normalization is suﬃcient. Flux calibration is useful for analysis of the stellar
component of the spectrum, where the detailed ﬂux of the continuum contains information
on the cluster’s stellar population.

The stellar component of the globular cluster is ﬁtted with a synthetic stellar spectrum
as described in Section 2.2.6. The best ﬁt model is used to account for and remove any eﬀect
of the overall stellar population of the globular cluster on the measurement of the emission
line equivalent width in Section 2.3.1. The [O III] emission line complex is then analyzed for
variability in Section 2.3 and the velocity proﬁle is modeled in Section 2.4 to investigate the
geometry of the emitting region.
16

2.2.6

Stellar Component Model

The observed spectra presented in this work are a superposition of the emission line system
of interest and the emission from stars of the host globular cluster. In order to analyze the
emission line system we ﬁrst ﬁt and subtract the stellar component. In ﬁtting the stellar
component we follow the prescription of Koleva et al. (2008). A grid of synthetic spectra
was created using the Pegas`-HR code of Le Borgne et al. (2004) and the Elodie 3.1 spectral
e
library (Prugniel et al., 2007). The models assumed a Salpeter initial mass function (IMF)
and a single epoch of star formation in the cluster. The synthetic spectra were ﬁt using the
WHT data due to its large wavelength coverage and moderate instrumental resolution. The
WHT data were masked to exclude the region of [O III]λλ4959, 5007 and Hβ to eliminate
any eﬀects of emission lines. The data were then ﬁtted using the grid of spectra which
yielded a best ﬁt model with an age of 12 Gyr and an initial [Fe/H] of -1.2. In order to test
this model ﬁtting scheme an independent set of synthetic stellar population models from
Vazdekis et al. (2007) were re-sampled and injected with noise to match our data, and then
ﬁt using the method described above. This test yielded estimates of the uncertainty in the
stellar component parameters contributed by the model ﬁtting scheme of ±1.5 Gyr in age
and ±0.05 dex in metallicity. Given the uncertainties in the stellar population models and
the assumptions used in the generation of the synthetic spectra, we adopt total uncertainties
of age ±2 Gyr and metallicity ±0.1 dex. Repeating the stellar component ﬁtting while
adopting an IMF for Kroupa (2001) produces ﬁt stellar parameters of 14 Gyr and [Fe/H]
of -1.1. Both these parameters are within the estimated uncertainties of the modeling and
ﬁtting procedure, suggesting the eﬀect of IMF selection on the resulting stellar component
model is minimal for this study. In Maccarone et al. (2007) a value of [Fe/H]=−1.7 ± 0.2
17

and an old stellar population age were inferred from the optical colors of Rhode & Zepf
(2001). These parameters are in reasonable agreement with the values we ﬁnd here. A
more detailed analysis of the stellar component of the globular cluster are presented in the
following chapter.
The stellar population parameters determined above were used to construct synthetic
spectra to match the wavelength coverage and spectral resolution of each observation. These
synthetic spectra were then used to remove the stellar component from each observation.

2.3

VARIABILITY

A careful study of the variability of the [O III] complex equivalent width and velocity structure, when combined with the x-ray variability timescales, potentially provides information
on the size and density structure of the emitting region. By examining the equivalent width
of the [O III] over various timescales it is possible to constrain the ionizing photon path
length and estimate the size of the emitting region. The emission line velocity structure,
coupled with any variability in X-ray data, may also help constrain other physical properties
of the emitting material.

2.3.1

Equivalent Width Variability

Line equivalent widths were measured by direct integration under the line proﬁle and bounded
by a linear ﬁt to the local continuum. The [O III]λλ4959,5007 emission is observed as a
blended feature, due to the very broad width of each [O III] line, as seen in Figure 2.1. The
[O III]λλ4959,5007 emission lines and the stellar absorption of the globular cluster RZ2109
are also found to have a redshift of 1475 km s−1 (Zepf et al., 2007). Therefore, the equiv18

3.0
3.0
2.5

2.5

Normalized Flux

2.0
1.5

2.0

1.0
0.5

1.5

4900

4950

5000

5050

5100

1.0

0.5

4600

4800

5000
5200
5400
5600
˚)
Observed Wavelength (A

5800

6000

Figure 2.1 Gemini GMOS spectrum of black hole hosting globular cluster RZ2109. The
inset displays the [O III]λλ4959,5007 emission structure which dominates the spectrum.
The spectrum has been continuum ﬁt and normalized, and smoothed with a three pixel
boxcar function.

19

alent width of the entire complex was measured between the observed wavelengths of 4964
˚ and 5058 ˚, which are symmetric in velocity space about the midpoint of the redshifted
A
A
broad [O III]λλ4959,5007 emission and beyond which the emission ﬂux is indistinguishable
from zero. The equivalent widths of the individual λ5007 and λ4959 lines were then set
by the ratio established by their respective transition probabilities. Identical measurements
were performed on the stellar component models and the resulting values subtracted from
the data measurements to yield equivalent widths of the emission component. Uncertainty
estimates for the equivalent widths were generated by considering a systematic contribution
(instrumental noise, poor cosmic ray subtraction, etc.) and a photon statistic contribution.
The systematic contribution was estimated by measuring the per pixel root-mean-squared
(RMS) deviation from the continuum of two regions located to the red and blue side of the
emission feature which were free of strong stellar absorption features. The two uncertainty
contributions were added by quadrature to produce the given values.

Table 2.1 displays the measured equivalent widths of the [O III]λ5007 for each observation
along with the date of observation. The uncertainty-weighted mean of all observations is
33.4 ± 0.3 ˚ with little change observed over the course of our observations. For the longest
A
time baseline between the ﬁrst observation with the Keck telescope to the last with GeminiSouth, we ﬁnd very close agreement in the [O III]λ5007 emission, with a a diﬀerence of only
∼ 9 ± 2% between these observations separated by 467 days. The intervening WHT and
SOAR data have larger uncertainties, but no observation deviates strongly from the mean
[O III] equivalent width, with the WHT observation being the most oﬀset with a formal
diﬀerence of 27.5 ± 7.5% from the weighted mean.
20

Table 2.1. [O III]λ5007 Equivalent Widths
eW (˚)
A

Observation
Keck
WHT
SOAR
GMOS

2.3.2

30.9 ± 0.7
42.6 ± 2.5
26.8 ± 8.0
34.0 ± 0.4

Date (days)
0.0
20.0
434.1
467.0

Line Proﬁle Variability

It is possible to examine the internight variability of the [O III] complex velocity proﬁle
using the Gemini spectra co-added for each of the three consecutive nights of observation.
In Figure 2.2 the three two hour exposures are displayed. A visual inspection of the three
spectra does not reveal any obvious variability over the region displayed. To test the line
proﬁle variability each spectrum was divided into 5 ˚ bins and subtracted from the other
A
two binned spectra. The diﬀerences in the bins in the [O III] region were then calculated
in terms of the RMS of the diﬀerences in bins not containing [O III] emission. Of the 57
comparison bins in the [O III] region, 17 displayed diﬀerences larger than 1σ, two larger
than 2σ and none larger than 3σ. For these diﬀerences, seven of the 17 1σ deviations were
above the continuum and ten were below it. These numbers are consistent with the RMS
statistical expectations, indicating no evidence for variability in the velocity structure over
the three nights observed.
A similar approach may be used to probe a longer temporal scale by a comparison of
the Keck and Gemini spectra. For this analysis the six hours of Gemini integrations are coadded and convolved with the Keck instrumental resolution. The Keck and convolved Gemini
spectra are displayed in Figure 2.3. A visual inspection reveals that when the resolution of
21

7

Normalized Flux

6
5
4
3
2
1

4800

4850

4900

4950
5000
5050
Wavelength (˚)
A

5100

5150

5200

Figure 2.2 Gemini spectra internight variability. The lower (March 28, 2009), middle (March
29, 2009), and upper (March 30, 2009) spectra have been smoothed with a three pixel boxcar
function and oﬀset on the ﬂux axis for legibility. The dashed lines give the level of the
normalized continuum for reference.

22

2.5

Normalized Flux

2.0

1.5

1.0

0.5
4800

4850

4900

4950
5000
5050
Wavelength (˚)
A

5100

5150

5200

Figure 2.3 Keck and Gemini spectra comparison. The Gemini spectra (dotted line) have
been convolved with the instrumental resolution of the Keck spectra (solid line.) The Keck
observation preceded the Gemini observation by 467 days.For interpretation of the references
to color in this and all other ﬁgures, the reader is referred to the electronic version of this
dissertation.

23

the two spectra are matched, the observations are consistent. A similar analysis as performed
on the Gemini individual night spectra yields 11 bins in the [O III] emitting region, four of
which have diﬀerences greater than 1σ, one greater than 2σ and none greater than 3σ. These
results are again consistent with the RMS statistical expectations. The 10 ˚ bin with the
A
greatest diﬀerence of 2.4 σ is centered at 5063 ˚, on the extreme red side of the emission
A
complex. By eye it is possible to pick out a small excess of emission in what appears to be
a ’secondary peak’ on the blue side of both the [O III]λ4959 and [O III]λ5007 structures in
the Keck spectrum. However when the λ4959 and λ5007 lines are transformed to velocity
space these ’secondary peaks’ are separated by 330 km s−1 , suggesting they are not real
structures.

2.4

GEOMETRIC MODEL

An initial qualitative inspection of the [O III]λλ4959, 5007 velocity structure line proﬁle
reveals that the complex may be divided into a λ4959 line and a λ5007 line shown in Figure
2.4. Theoretical considerations dictate that these two lines should display identical line
proﬁle shapes and diﬀer only in their total ﬂux level, the ratio of which is set by their
transition probabilities. The data are in good agreement with this expectation. The two
main features of the [O III] line proﬁles are strong broad emission spanning between ∼300–
1300 km s−1 , and a central, narrower peak. At larger velocities the emission falls oﬀ rapidly
between ∼1300–1700 km s−1 . While the velocity structure appears symmetric about line
center, there appears to be a ﬂux level asymmetry between the blue and red sides of the line
proﬁle.
In an eﬀort to quantify the observed emission line proﬁle, we ﬁt a simple geometric model
24

Normalized Flux
−1500

−1000

0
500
−500
Velocity (km/s)

1000

1500

Figure 2.4 Gemini spectra [O III] velocity structure. The top ﬁgure displays the line proﬁle
for [O III]λ5007, and the lower panel [O III]λ4959. The two lines have been separated by a
cut at an observed wavelength of λ5006. As a result there may be a small degree of blending
on the extreme blue side of the [O III]λ5007 and the extreme red side of [O III]λ4959. The
dashed lines give the continuum level for reference.

25

to the [O III]λ5007 line proﬁle of the continuum normalized Gemini data. The model consists
of two components; a bipolar conical outﬂow and a lower velocity width Gaussian component.
The bipolar conical outﬂow produces the broad emission component, with the lower velocity
width peak contributed by the Gaussian. This two component model is necessary as no
model with only a bipolar conical component or only a Gaussian component was able to
simultaneously reproduce the ﬂat broad component, and the narrower peak. Fitting the
simple geometric model to the observations allows us to describe the emission line structure
in terms of the model parameters: the velocity of the outﬂowing gas, the opening angle of
the cones, the angle of inclination of the cone’s axis relative to the observer, and the full
width half maximum (FWHM) of the Gaussian component.

The resulting best ﬁt model is displayed along with the Gemini data in Figure 2.5. This
model has an outﬂow velocity of 1600+190 km s−1 , an opening angle of 70 ± 8 degrees, and
−90
an inclination angle of 65+3 degrees relative to the line of sight. The uncertainty values on
−6
these parameters indicate the value at which a given parameter produces a 3 σ deviation
of the model velocity proﬁle from the Gemini data, with the other two parameters ﬁxed to
the best ﬁt values. It is important to note that the model parameters are not completely
independent, and so these uncertainties do not demarcate a region of “acceptable” ﬁts in
parameter space. For reference we include a set of cross-sectional uncertainty contours about
the best ﬁt model parameters in ﬁgure 2.6. The Gaussian component of the model has a
FWHM of ∼ 310 km s−1 . In this model 81 percent of the ﬂux is contributed by the bipolar
conical outﬂow, and the remaining 19 percent by the lower velocity width Gaussian structure.
The ﬂux ratio of the receding outﬂow to the approaching outﬂow is 1.4. This simple model
provides an excellent ﬁt to the [O III]λ5007 line proﬁle out to ∼1300 km s−1 from line center.
26

3.0

Normalized Flux

2.5

2.0

1.5

1.0
−1500

−1000

−500

0
500
Velocity (km/s)

1000

1500

2000

Figure 2.5 Gemini spectra with geometric model. The solid black line shows the Gemini
spectra, the dotted line displays the best ﬁt bipolar conical outﬂow component of the geometric model, the dashed line the Gaussian component of the model, and the solid gray line
(red in the electronic version) the full geometric model.

27

90
80
70
60
50
40

Velocity (km/s)

1800

2000

1900

1800

1700

1600

1300

20
1900

1500

30
1400

Opening Angle (deg)

100

Velocity (km/s)

1700
1600
1500
1400
1300
1200

45

50
55
60
65
70
Inclination Angle (deg)

Figure 2.6 Geometric model uncertainty contours. The top left plot displays the uncertainty
contours in opening and inclination angles with a ﬁxed velocity of 1600 km s−1 . The top right
plot give the uncertainty contours in opening angle and velocity about a ﬁxed inclination of
65 deg. The bottom plot shows the uncertainty contours in velocity and inclination angle
about a ﬁxed opening angle of 70 deg. Contours in each plot give the 2, 3, 4, and 5 sigma
deviations from the Gemini data. The plotted cross shows the position of the best ﬁt value.

Above this velocity the data show excess emission relative to the model both at the red and
blue extremes.
The model assumes an optically thin emitting region, as appropriate for [O III], and a
uniform emission density. The uniform emission density is adopted to simplify the calculation, but provided there is no strong density dependence with the angular separation from
the central axis, any radial dependence of the emission density will not change the velocity
structure of the model line. The bulk ﬂow velocity is taken to be constant over the outﬂow.
28

This assumption, while assuredly an oversimpliﬁcation, provides a reasonable description
of the available data. It is worth noting that in this formulation of the geometric model,
the Gaussian component does not have an explicit physical analog. A number of physical
structures could produce a Gaussian velocity proﬁle and we will consider a few possibilities
in later sections.

Given the simplicity of the high velocity width bipolar conical outﬂow plus lower velocity
width Gaussian model, we do not place great emphasis on the detailed parameters of the
model ﬁt. Broadly, the large velocity width requires a signiﬁcant outﬂow velocity. The ﬂux
level asymmetry of red and blue high velocity structures requires a non-uniform outﬂow. A
ﬂat velocity proﬁle, as observed in the high velocity width component, may also be produced
by a uniform spherical outﬂow or approximated by a radially directed disk wind. However,
the ﬂux asymmetry of approaching and receding outﬂows observed favors the spatially discrete structures of the bipolar conical outﬂow considered above. Due to this asymmetry, the
best ﬁt proﬁle of a spherical outﬂow deviates from the Gemini data at the 9 sigma level. A
spherical shell will produce an identical proﬁle to a ﬁlled sphere because, as noted above, the
velocity proﬁle of any outﬂow whose velocity is independent of radius, is also independent
of the radial density proﬁle. The large opening angle indicates that the outﬂow that ﬁts
best in this model is not from a narrowly collimated jet, but rather a broader outﬂow. Any
model which includes the detailed hydrodynamics of the outﬂowing gas may deviate from
the parameters generated by the purely geometric ﬁt considered here. As such we present
the model parameters as descriptive and useful for consideration of the general properties of
the emitting region.
29

2.5
2.5.1

DISCUSSION
Variability

There are two major modes by which the observed lack of strong variability in the [O III]
emission might be produced in an accreting black hole system; a steady accretion mode, or an
emitting system in which the crossing time for ionizing photons is large in comparison with
the variability timescale of the ionizing source. In the case of a variable ionizing source the
emission variability may also contain information on the density distribution of the emitting
gas. The x-ray data described in Shih et al. (2008) span nine years and display a consistent
count rate between observations. The four observations described in the present work also
display consistent [O III]λ5007 equivalent widths between observations. For reference the
light crossing path bounded by the optical observations presented here is 0.4 pc. Both
observations are consistent with a long lived steady accretion rate. The sparse temporal
sampling of the data, however, does not allow tight constraints on shorter term ﬂuctuations
in either the X-ray or [O III] emission. Future x-ray observations of the RZ2109 x-ray source,
XMMU 1229397+07533, with ﬁner temporal resolution could be combined with the [O III]
variability limits presented here to construct limits on the spatial scale of the emitting region.
Likewise variability in the [O III] emission under constant x-ray ﬂux could yield information
on the outﬂow rate and density distribution.
The absence of strong variability does allow for the elimination of one class of models
as the source of the observed [O III] emission. Novae shells in the local Milky War are
a known source of [O III]λ5007 emission (Downes et al., 2001). The very brightest shell
novae reach L([O III]λ5007) ∼ 1037 ergs and have velocity widths on the order of those
30

observed in RZ2109 (Downes et al., 2001; Iijima & Naito, 2011, for example). In contrast to
the observations presented here, however, the [O III] luminosity of these sources decay by
several orders of magnitude and velocity proﬁle vary on timescales of days to weeks.

2.5.2

Structure of Emission Region

It is somewhat remarkable that the geometric model presented in Section 2.4 ﬁts the [O III]
line proﬁle as well at it does with simplistic assumptions. There are, of course, a number
of questions either not addressed or raised by the model, including the spatial scale of
the emitting system, the observed ﬂux asymmetry of the red and blue high velocity width
component, the failure of the model at the extreme high velocity wings, and the nature of
the low velocity Gaussian component.
The geometric model as presented is free of any spatial scaling. We must look to other
means in order to place limits on the size of the outﬂow region. A ﬁrm lower limit may be set
using the total emission from the high velocity width outﬂow component of the [O III]λ5007
emission, adopting a pure [O III] gas at critical density for [O III]λ5007 at a temperature of
104 K (Osterbrock, 1989), and employing the geometry parameters found. With this minimal
constraint the observed emission could originate from a region with a minimum radius of
a few ×10−3 pc. The data do not provide a clear upper limit on the size of the emitting
region, and there is no reason given the available data that the size of the [O III]λ50007
region could not be as large as the ∼ 10 pc extent of the globular cluster itself.
Asymmetries in the bipolar outﬂows are a well known feature of accreting black hole
systems. When a ﬂux asymmetry is attributed to a diﬀerential obscuration the usual pattern
observed is a decrement in the ﬂux of the receding ﬂow produced by the additional obscuring
31

material along the greater line of sight distance (van Groningen, 1987; Peterson et al., 2000).
However, we observe the opposite eﬀect in the RZ2109, with the receding [O III] emission line
40 percent more luminous than the approaching ﬂow. If diﬀerential obscuration contributes
to the observed ﬂux asymmetry, the obscuring media must be local to the cluster system,
such as dust-bearing gas at large radius resulting from an earlier outﬂow of the system or
interaction with a asymmetric ISM. The mechanism by which such a local obscuring media
may be produced is unclear.
A second known source of ﬂux asymmetry in outﬂows is Doppler boosting (Mirabel &
Rodr´
ıguez, 1999; Fender, 2006). In Doppler boosting, however, the ﬂux of the approaching outﬂow is ampliﬁed relative to the receding outﬂow, again opposite of the asymmetry
observed in RZ2109. The diﬀerent line of sight travel times of outﬂows toward and away
from the observer can also produce asymmetric ﬂuxes if coupled with a variable ionizing
source. However, the ﬂux asymmetry is the same in both our Keck and Gemini observations
separated by more than a year, which would be very diﬃcult to understand in such a scenario. It therefore seems most likely that the somewhat greater ﬂux on the red side of the
[O III]λ4959,λ5007 line proﬁle relative to the blue side is a result of an intrinsic asymmetry in
the source’s emission density. Such an asymmetry may be attributable to either a diﬀerence
in the mass of each outﬂow or the ionizing ﬂux incident on them.
An expanding gas distribution such as nova shell or other remnant produced by late stage
stellar evolution and ionized by an external source may produce a ﬂux asymmetry across
the velocity proﬁle provided the gas is optically thick. Such an arrangement would allow
for a ﬂux excess of the receding side of the velocity proﬁle relative to the approaching if
the media was positioned between the observer and the ionizing source. For example, a late
32

stage nova shell modeled with a uniform density shell illuminated by an external x-ray source
produces an asymmetric velocity proﬁle with a continuous blue to red ﬂux gradient whose
slope is set by the position of the external x-ray source. However, this type of ﬂux asymmetry
fails to provide an adequate description of ﬂat topped step function proﬁle observed in the
[O III]λ4959,λ5007 line proﬁles high velocity width component.
The failure of the model at the most extreme velocities, seen in velocities in excess
of ±1400 km s−1 in Figure 2.5, is most likely attributable to the simplifying assumption
of constant bulk velocity in the geometric model. If this is the case and the emission in
these extreme velocity wings is produced by the outﬂow closest to the source, prior to any
deceleration, with more sophisticated modeling techniques the extreme velocity wings may
provide another constraint on the spatial scale of the system.
While a model with a strong bipolar outﬂow can naturally account for broad [O III]λ4959,λ5007
emission, the origin of the narrower Gaussian component is less clear. The observed FWHM
of this lower velocity width Gaussian peak is three times the FWHM of the spectral resolution of the Gemini spectrum. Thus, the observed strongly peaked [O III]λ5007 line proﬁle
is representative of its intrinsic shape. One implication of the strongly peaked line proﬁle is
that rotation is ruled out for the origin of the observed width of the line. It is worth noting
that forbidden lines such as [O III] are optically thin, and the observed proﬁle reﬂects the
intrinsic one, unlike the much more complex case of optically thick permitted lines such as
Balmer emission lines in AGN. Therefore, for optically thin [O III], rotation naturally produces a double-peaked or ﬂat-topped velocity proﬁle (e.g. Clark et al., 1979), and fails badly
to match the centrally peaked proﬁle seen in the [O III] emission. For example, the speciﬁc
case of a rotating Keplerian disk with an inverse square law emission density proﬁle has a
33

velocity proﬁle which deviates from the observed lower velocity width [O III] component at
the 7σ level in a one spectral resolution element bin about line center. In fact, instead of
showing a double-horned proﬁle or very ﬂat-topped proﬁle typical of rotation, the observed
proﬁle is more centrally peaked than a Gaussian at about the 3σ level. Therefore, rotation is
ruled out as the source of the velocity width of the lower velocity width Gaussian component,
contrary to the assumption of some modeling (Porter, 2010).
The luminosity of the L([O III])λ5007 at a given velocity can also be used to determine
whether the velocity width of the line can be due to gravitational motions or not. Speciﬁcally,
because [O III]λ5007 is a forbidden line with a known critical density (Osterbrock, 1989),
we can determine the maximum possible L([O III])λ5007 within a given volume by adopting
a pure [O III] gas of uniform density at its critical density. We can also compute the largest
volume available suﬃciently close to the black hole to account for the observed velocity
width due to gravitational motions about the black hole. Zepf et al. (2008) carried out this
calculation for the broad velocity component of the [O III]λ5007, and showed it is many
orders of magnitude stronger than can be accounted for with gravitational motions, and to
thus require strongly outﬂowing material.
Here we apply this same calculation to the lower velocity width Gaussian component
which is seen much more clearly in our new Gemini data with its greater spectral resolution
and higher signal-to-noise. Adopting a pure [O III] gas of uniform density at its critical
density (Osterbrock, 1989) and a canonical black hole mass of 10 M⊙ , we ﬁnd the maximum
luminosity in a one resolution element bin -170 km s−1 from line center is of order 1029 ergs
s−1 . In contrast, the observed luminosity at this velocity is 4.2 × 1035 ergs s−1 , and thus
gravitational motions also fail to account for the velocity width of the lower velocity width
34

Gaussian peak.
It is worth noting that the above arguments do not strictly rule out rotation of the
emission line region, they do, however, require that the observed line widths be produced
by some mechanism other than rotation. A scenario where the emitting region rotates in
the plane of the sky, or with speciﬁc arrangements of absorbing systems may be invoked,
but in these cases the line width and proﬁle would be independent of any rotation. Because
the emitting region is taken to be optically thin to the forbidden line emission observed, self
absorption or intermediate line absorption systems are unlikely. Gray body or broad band
absorption/scattering via an obscuring media is possible, however the speciﬁc distribution of
the material necessary to transform a rotational line proﬁle to that observed (a preferential
and symmetric obscuration of the emitting gas moving parallel/anti-parallel with the line of
sight) is diﬃcult at best.
Although the speciﬁc calculation above is for a rotating sphere for simplicity, adopting
a dynamically hot model for the emitting material will change the gravitational motions
prediction far less than the ﬁve or more orders of magnitude diﬀerence between it and the
data. The calculation also adopts a stellar mass for the black hole in the globular cluster,
based on results in Z08. In that work, Z08 used the same type of volume calculation as
above for the broad component of the [O III]λ5007 emission, and showed that it is not due
to gravitational motions unless the black hole is nearly as massive as the ∼ 106 M⊙ cluster
itself (for a more detailed treatment see also Porter, 2010). The broad component then
must be due to a strong outﬂow. Z08 suggest that observational evidence and theoretical
calculations (Proga, 2007, and references therein) indicate that strong outﬂows are associated
with high (L/LEdd ) systems, which given the observed LX , implies a stellar mass for the
35

black hole in RZ2109.
An alternative idea, that the source could be a tidal disruption or detonation of a white
dwarf by an intermediate-mass black hole, fails on several grounds (e.g. Irwin et al., 2010). A
disruption is unable to account for either the high L([O III]λ5007) over an extended period
of time or the long duration of its high X-ray ﬂux from at least the early 1990s ROSAT
era through 2004 (Maccarone et al., 2007; Shih et al., 2008). A detonation may account for
the high velocity outﬂow, but does not provide an explanation for why only oxygen lines
are seen in the spectrum when the detonation is powered by the fusion of oxygen into iron.
Furthermore, neither a disruption or detonation is able to account for the more than a decade
of consistently high LX followed by the recently observed steep decline of LX by more than
an order of magnitude (Maccarone et al., 2010).
With the current observations we have limited ability to address the energetics of the
outﬂow. Since we have no direct measurement of the temperature, we can estimate the mass
of [O III] in the emitting region ﬁnding using a temperature typical of [O III] emission line.
With an assumption of T = 104 K we ﬁnd an [O III] mass of 7.3 × 10−5 M⊙ . For temperatures from 5 × 103 –106 K the mass ranges from 1 × 10−3 to 4 × 10−6 M⊙ . Unfortunately
since we have little to constrain either the composition of gas, beyond that it is likely hydrogen depleted, or the ionization fraction as no [O II] lines are observed above the detection
threshold, we are unable to provide a concrete estimate of the total gas mass of the emitting
region. For a pure oxygen gas the total mass will be within a factor of a few greater than
the [O III] mass, while a solar composition would imply a total mass roughly two orders of
magnitude greater. Using the 1600 km s−1 velocity of the higher velocity component found
in Section 2.4 the kinetic energy of the [O III] is then 1045 erg. This energy is 8 orders of
36

magnitude smaller than the available total accretion energy, assuming an accretion eﬃciency
of 10 percent and a 1 M⊙ donor. If a solar composition is assumed for the outﬂowing gas and
all oxygen being doubly ionized the systems kinetic energy is 6 orders of magnitude smaller
than the total accretion energy. It should be noted that the kinetic energy given here is
for the total observed outﬂow; it is unknown what fraction of the total accretion event the
current outﬂow represents.

2.6

SUMMARY

In this work we have presented a study of the equivalent width and velocity proﬁle of the
[O III]λλ4959, 5007 emission line complex associated with the black hole X-ray source in the
extragalactic globular cluster RZ2109. We have found little variation in the [O III]λ5007
emission in our multiple observations over a time span of 467 days. The ﬁrst and last
observations have a formal diﬀerence of 9 ± 2% with no evidence for diﬀerent velocities,
and our upper limit in the variation among all our observations is 27%. The [O III]λ5007
emission line is comprised of two components; a lower velocity width component which is well
characterized by a Gaussian with FWHM 310 km s−1 , and high velocity width component
with a width of 3200 km s−1 which we describe using a simple bipolar conical outﬂow
model. The velocity width of the Gaussian component cannot be due to rotation, and both
components appear to require non-gravitational motions in order to provide suﬃcient volume
to produce the observed high luminosities in the [O III]λ5007 line.

ACKNOWLEDGEMENTS
MMS and SEZ wish to acknowledge support from NSF grants AST-0406891 and AST37

0807557. AK thanks NASA for support under Chandra grant GO0-11111A. KLR acknowledges support from NSF Faculty Early Career Development grant AST-0847109. We thank
the anonymous referee for the questions and critique which helped to strengthen the paper.
We also thank Jack Baldwin for helpful discussion.

38

BIBLIOGRAPHY

39

BIBLIOGRAPHY

Casares, J., Steeghs, D., Hynes, R. I., Charles, P. A., & O’Brien, K. 2003, ApJ, 590, 1041
Clark, F. O., Martin, H. M., & Biretta, J. A. 1979, ApJ, 232, 624
Clemens, J. C., Crain, J. A., & Anderson, R. 2004, Proc. SPIE, 5492, 331
Cooke, R., Bland-Hawthorn, J., Sharp, R., & Kuncic, Z. 2008, ApJ, 687, L29
Downes, R. A., Duerbeck, H. W., & Delahodde, C. E. 2001, Journal of Astronomical Data,
7, 6
Eggleton, P. P. 1983, ApJ, 268, 368
Fender, R. 2006, Compact stellar X-ray sources, 381
Gnedin, O. Y., Maccarone, T. J., Psaltis, D., & Zepf, S. E. 2009, ApJ, 705, L168
Hook, I. M., Jørgensen, I., Allington-Smith, J. R., Davies, R. L., Metcalfe, N., Murowinski,
R. G., & Crampton, D. 2004, PASP, 116, 425
Iijima, T., & Naito, H. 2011, A&A, 526, A73
Irwin, J. A., Brink, T. G., Bregman, J. N., & Roberts, T. P. 2010, ApJ, 712, L1
Ivanova, N., Chaichenets, S., Fregeau, J., Heinke, C.O., Lombardi, J.C., & Woods, T.E.
2010, ApJ, 717, 948
Kaaret, P., & Corbel, S. 2009, ApJ, 697, 950
40

Koleva, M., Prugniel, P., Ocvirk, P., Le Borgne, D., & Soubiran, C. 2008, MNRAS, 385,
1998
Kroupa, P. 2001, MNRAS, 322, 231
Le Borgne, D., Rocca-Volmerange, B., Prugniel, P., Lan¸on, A., Fioc, M., & Soubiran, C.
c
2004, A&A, 425, 881
Macri, L. M., et al. 1999, ApJ, 521, 155
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2007, Nature, 445, 183
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2010, MNRAS, 409, L84
Mirabel, I. F., & Rodr´
ıguez, L. F. 1999, ARA&A, 37, 409
Nelemans, G., Jonker, P. G., & Steeghs, D. 2006, MNRAS, 370, 255
Oke, J. B., et al. 1995, PASP, 107, 375
Osterbrock, D. E. 1989, Astrophysics of Gaseous Nebulae and Active Galactic Nuclei (Mill
Valley, CA: Univ. Science Books)
Peterson, B. M., et al. 2000, ApJ, 542, 161
Porter, R.L. 2010, MNRAS, in press
Proga, D. 2007, The Central Engine of Active Galactic Nuclei, 373, 267
Prugniel, P., Soubiran, C., Koleva, M., & Le Borgne, D. 2007, arXiv:astro-ph/0703658
Rhode, K. L., & Zepf, S. E. 2001, AJ, 121, 210
Russell, D. M., Fender, R. P., Gallo, E., & Kaiser, C. R. 2007, MNRAS, 376, 1341
Shih, I. C., Maccarone, T. J., Kundu, A., & Zepf, S. E. 2008, MNRAS, 386, 2075
Remillard, R. A., & McClintock, J. E. 2006, ARA&A, 44, 49
41

van Groningen, E. 1987, A&A, 186, 103
Vazdekis, A., et al. 2007, IAU Symposium, 241, 133
Werner, K., Nagel, T., Rauch, T., Hammer, N. J., & Dreizler, S. 2006, A&A, 450, 725
Zepf, S. E., Maccarone, T. J., Bergond, G., Kundu, A., Rhode, K. L., & Salzer, J. J. 2007,
ApJ, 669, L69
Zepf, S. E., et al. 2008, ApJ, 683, L139

42

Chapter 3
Composition of an Emission Line
System in Black Hole Host Globular
Cluster RZ2109
Matthew M. Steele 1 , Stephen E. Zepf 1 ,Arunav Kundu 1 2 , Thomas J. Maccarone3 ,
Katherine L. Rhode 4 , and John J. Salzer 4

We present an analysis of the elemental composition of the emission line system associated
with the black hole hosting globular cluster RZ2109 located in NGC4472. From medium
resolution GMOS optical spectroscopy we ﬁnd a [O III]λ5007/Hβ emission line ratio of 106
for a 3200 km s−1 measurement aperture covering the full velocity width of the [O III]λ5007
1 Department of Physics & Astronomy, Michigan State University, East Lansing, MI

48824; e-mail: steele24@msu.edu
2 Eureka Scientiﬁc, 2452 Delmer Street Suite 100, Oakland, CA 94602-3017
3 School of Physics & Astronomy, University of Southampton, Southampton, Hampshire
S017 1BJ, UK
4 Department of Astronomy, Indiana University, Bloomington, IN 47405
43

line, with a 95% conﬁdence level lower and upper limits of [O III]λ5007/Hβ > 35.7 and
< −110 (Hβ absorption). For a narrower 600 km s−1 aperture covering the highest luminosity
velocity structure in the line complex, we ﬁnd [O III]λ5007/Hβ = 62, with corresponding 95%
conﬁdence lower and upper limits of > 30.2 and < −364. The measured [O III]λ5007/Hβ
ratios are signiﬁcantly higher than can be produced in radiative models of the emission
line region with solar composition, and the conﬁdence interval limits exclude all but the
most extremely massive models. Therefore, we conclude that the region from which the
[O III]λ5007 emission originates must be hydrogen depleted relative to solar composition
gas. This ﬁnding is consistent with emission from an accretion powered outﬂow driven by a
hydrogen depleted donor star, such as a white dwarf, being accreted onto a black hole.

3.1

INTRODUCTION

The globular cluster RZ2109 located in the galaxy NG4472 is a known host of an accreting
black hole system, ﬁrst identiﬁed by (Maccarone et al., 2007). Along with the variable x-ray
source indicative of the accreting black hole, RZ2109 has been observed to host a broad and
luminous [O III]λλ4959,5007 emission line complex thought to correspond to an accretion
powered outﬂow driven from the cluster’s black hole (Zepf et al., 2007, 2008; Steele et al.,
2011).
The RZ2109 emission line structure is unusual in many respects. The lines are very
broad with a width of 3200 km s−1 , about a factor of 100 larger than the escape velocity
of the globular cluster in which it is located. The line velocity proﬁles also have a complex
shape and appear to contain two distinct velocity structures (Steele et al., 2011). Moreover, [O III]λλ4959,5007 are the only emission lines apparent even in a fairly deep spectrum
44

(Zepf et al., 2008). Gnedin et al. (2009) suggested that the lack of Balmer line emission, in
particular, may be indicative of a hydrogen poor donor star such as a white dwarf. This possibility is signiﬁcant as while BH-WD binaries are an expected end stage of stellar evolution,
especially in globular clusters, none has yet been positively identiﬁed (Ivanova, 2011).
A common diagnostic for the study of emission line regions is the [O III]λ5007/Hβ emission line ratio. Previous work by Zepf et al. (2008) noted that [O III]λ5007/Hβ ratio appaeared to be at least 30 based on a low resolution optical spectrum of the RZ2109 cluster.
For reference other common astrophysical [O III] emission line production sites include active
galactic nuclei with [O III]λ5007/Hβ ratios of order unity (Sarzi et al., 2006), Milky Way
planetary nebulae with mean line ratios of approximately 15 (Stanghellini et al., 2003), and
a few extremely hydrogen poor planetary nebulae with line ratio of 20 (M´ndez et al., 2005;
e
Larsen, 2008). By comparison to these sources is seems plausible that the RZ2109 emission
line site would be hydrogen depleted, however the [O III] production sites and mechanisms
can be very diﬀerent across astrophysical object classes and it is not immediately obvious
how apt such comparisons are.
In this work we address the composition of the RZ2109 emission line region in order to
more tightly constrain models of the black hole/donor star binary in the globular cluster.
For this we analyze higher resolution spectra of RZ2109 at better signal-to-noise than have
previously been applied to the question of the composition of the RZ2109 emission line source,
described in section 3.2. In section 3.3.1 we model the stellar component of the RZ2109
spectrum in order to remove any stellar contributions from the emission line measurements
presented in section 3.3.2. To understand the range of [O III]λ5007/Hβ emission line ratios
possible for diﬀerent conﬁgurations of a solar composition emission line region, a suite of
45

radiative models are considered in 3.3.3.

3.2

OBSERVATIONS AND DATA REDUCTION

The optical spectra used in this study were collected using the Gemini Multi Object Spectrograph (GMOS) on the Gemini South Telescope. These data, ﬁrst presented in Steele et
al. (2011), were taken on the consecutive nights of UT 2009 March 28–29 (program GS2009A-Q-1). The medium resolution spectra (R ∼ 2400) have a wavelength coverage from
4445–6023 ˚. The data were reduced using the longslit tools in the Gemini IRAF package.
A
One dimensional spectra were extracted from the processed spectra and, following tests to
ensure the emission line object was not varying between nights (Steele et al., 2011), co-added
to increase the signal-to-noise ratio.
The resulting reduced spectrum is a superposition of the emission line source of interest
and the stellar component of the host globular cluster. In order to examine the emission line
source in isolation, the stellar component must be modeled and removed.

3.3
3.3.1

ANALYSIS AND RESULTS
Stellar Component Modeling

The stellar modeling was performed in a process similar to that described in Steele et al.
(2011). The RZ2109 Gemini spectrum was ﬁt to a grid of continuum normalized synthetic
spectra constructed using the Pegas´-HR code of Le Borgne et al. (2004) and Elodie 3.1
e
(Prugniel et al., 2007) stellar library. A Kroupa IMF (Kroupa, 2001) was adopted along with
a single epoch of star formation to construct the synthetic stellar population grid using cluster
46

3.0
1.4
1.3
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5

Normalized Flux

2.5

2.0

1.5

4840 4860 4880 4900 4920 4940

1.0

0.5

4600

4800

5000 5200 5400 5600
Observed Wavelength (˚)
A

5800

6000

6200

Figure 3.1 GMOS spectrum and the synthetic stellar component model. The GMOS observations of RZ2109 are shown using the black line. The gray line (red in the electronic
version) depicts the best ﬁt synthetic stellar model. The inset shows the wavelength region
surrounding Hβ in detail.

47

age and initial metallicity as free parameters. To avoid emission lines contaminating the ﬁt
with the inclusion of emission lines with non-stellar origins, the region between 4950-5060 ˚
A
in the observed frame was masked out to exclude contributions from Hβ and [O III]λ5007
to the χ2 minimization ﬁtting scheme. The resulting best ﬁt stellar model corresponded to
an age of 13.25 Gyr and a metallicity of [Fe/H] = -1.0.

The resulting best ﬁt synthetic stellar population model provides an excellent match to
the stellar component of the observed spectrum of RZ2109. Figure 3.1 provides a comparison
between the observed spectrum and the best ﬁt model, over the entire spectrum wavelength
range along with an inset showing the Hβ region in detail.

In order to determine the uncertainties in the ﬁtted parameters, simulated observations
were preformed by constructing a best ﬁt synthetic stellar component spectrum and injected
it with noise to match the signal-to-noise of the observations. The simulated observation was
then ﬁt to produce a new set of stellar parameters. This simulated observation process was
performed for 105 iterations with the resultant probability density distribution of the stellar
component parameter ﬁts shown in ﬁgure 3.2. From this exercise it is evident that the ﬁt is
well constrained in [Fe/H], while displaying a greater spread in the age parameter. The log
scale used for the stellar parameter density distribution in ﬁgure 3.2 may overemphasize the
spread in age; only 1 percent of the of the simulated observations had age determinations
below 5.75 Gyr and only 10 percent were below 8.50 Gyr. The adopted stellar parameters
and associated uncertainties (Age = 13.25 ± 1.00 Gyr, [Fe/H] = −1.0 ± 0.1) include 83
percent of all simulated observations in age and 98 percent in metallicity.
48

16
−0.5
14

−1.0
0 .0 0

12

10

−2.5

0 .5

Age (Gyr)

−2.0

−3.0

0

8

−3.5
6
1 .5

00

−4.0

0

−3.0

0

50

0

4

1.

2 .0

3 .0

2.

log Probability Density

−1.5

−4.5

3.50

−2.5

−2.0

−1.5
[Fe/H]

−1.0

−0.5

0.0

−5.0

Figure 3.2 Stellar Model Fitting Uncertainties and Hβ equivalent width contours. The
gray scale grid displays the probability density in stellar parameter space that the stellar
component of the RZ2109 spectrum would be determined to have a given set of parameters
as described in section 3.3.1. The contour lines show the measurement of the Hβ equivalent
width with a 600 km s−1 aperture for a given stellar component model relative to the best
ﬁt model, as described in section 3.3.2.

49

Normalized Flux
−3000

−2000

−1000

0
1000
Velocity (km/s)

2000

3000

Figure 3.3 Equivalent Width apertures. The upper panel displays the Hβ region of the
GMOS spectrum in velocity space, and the lower panel gives the [O III]λ5007 region. The
600 km s−1 aperture over which the equivalent width was measured is shaded with dark gray.
The 3200 km s−1 aperture is shown in light gray. The dashed line gives the continuum level.

50

Table 3.1. Emission line equivalent widths and Hβ ratios
Rest Wavelength (˚)
A

Aperture (km s−1 )

EW (˚)
A

X/Hβ

HI

4861

[O III]

5007

3200
600
3200
600

0.32 ± 0.32
0.21 ± 0.12 ± 0.04
33.82 ± 0.39
13.14 ± 0.10

1.0
1.0
105.7
61.6

Species

3.3.2

Equivalent Widths

The equivalent widths (EW) are measured by direct integration of the continuum normalized
spectrum bounded by a given velocity aperture. A second measurement is then performed
over the identical aperture on the stellar model and the value subtracted from the data
measurement in order to remove the inﬂuence of stellar absorption features. Given the
complexity of the [O III]λ5007 line proﬁle the speciﬁcation of the velocity aperture is nontrivial. Steele et al. (2011) ﬁnd that the [O III]λ5007 emission line in RZ2109 is well described
by two components; a broad component with a width of ∼ 3200 km s−1 and narrower
component with a FWHM of ∼ 300 km s−1 . Here we perform measurements for apertures
of two widths which correspond to the two components identiﬁed by Steele et al. (2011); a
broad 3200 km s−1 aperture and a 600 km s−1 aperture. For reference the apertures over
which the measurements are performed are plotted for the Hβ and [O III]λ5007 lines in
ﬁgure 3.3. Table 3.1 gives the resulting measurements for [O III]λ5007 and Hβ in both the
3200 and 600 km s−1 apertures.
The cited 1 σ uncertainties provide an estimate of the rms uncertainties derived from
the local continuum and Poisson statistic uncertainties for emission above the local continuum. To construct an estimate of the local continuum rms two regions, one blueward, one
51

redward, with a width equal to that of the measurement aperture were evaluated to ﬁnd
the mean expected deviations from the local continuum. The second uncertainty cited for
the 600 km s−1 aperture Hβ measurement reﬂects the uncertainty produced by the ﬁt of
the synthetic stellar population model. This uncertainty estimate was produced by injecting
noise consistent with that of the observations into a synthetic stellar model with parameters
matching the best ﬁt model, reﬁtting the resulting spectrum with a new stellar model using
the procedure described in section 3.3.1, and remeasuring the Hβ EW.
The contours plotted in ﬁgure 3.2 display the Hβ EW as a function of the stellar component parameters in units of a ratio to the 600 km s−1 aperture EW measurement cited
in 3.1. Considering the full stellar parameter space covered by the simulated observations
detailed in 3.3.1, the mean Hβ EW of the simulated observations is two percent greater than
the cited value, with 58 percent of the iterations falling within 10 percent and 89 percent of
iterations within 20 percent of the cited value. If we limit consideration to only those simulated observations located within the uncertainty bounds of the best ﬁt stellar parameters,
the mean Hβ EW is one percent less than the cited value, 66 percent of iterations are within
10 percent of the cited value and all simulated observations are within 20 percent. Therefore
an uncertainty in the 600 km s−1 aperture Hβ EW measurement of 20 percent is adopted.
In addition to the [O III]λ5007 and Hβ measurements described above, EW measurements
were performed for a selection of other emission lines which modeling suggested may be
present with EW a signiﬁcant fraction of that of Hβ. These additional measurements were
performed with a 600 km s−1 aperture on He IIλ4686 (EW= 0.02 ± 0.12), [Ar IV]λ4740
(EW= 0.07 ± 0.13), and [Fe VII]λ5721 (EW= 0.07 ± 0.08). It should be noted that other
than [O III]λ5007 and the 600 km s−1 aperture Hβ EW (which is a very weak detection), all
52

EW for both 3200 and 600 km s−1 apertures are consistent with non-detections.
As the two emission components are thought to be produced in spatially and geometrically distinct structures, the emission line ratios derived using the two apertures need not
be the same. Since the contributions of the broad and narrow emission components are
superimposed, both the 3200 km s−1 or 600 km s−1 aperture measurements reﬂect an EW
with contributions from both sources. In the 3200 km s−1 aperture the broad component
dominates the EW, while the narrow component dominates in the 600 km s−1 aperture measurements. For reference Steele et al. (2011) claim 81 percent of the total [O III]λ5007 ﬂux
is contributed by the the broad component and the remainder by the narrow component,
though it should be noted deconvolving these contribution is necessarily a model dependent
calculation.

3.3.3

Emission Line System Modeling

As a means of testing the possibility that the measured [O III]λ5007/Hβ emission line ratios
given above may be produced by a gas of solar composition, a series of radiative transfer
models were conducted. The models were executed using the Cloudy (version 08.01) spectral
synthesis code of Ferland et al. (1998).
The suite of models tested all had solar compositions and spherically symmetric shell gas
distributions characterized by an inner and outer radius . The gas was taken to be outﬂowing
at a velocity of 300 km s−1 at the inner radius. The gas is gradually accelerated by radiation
pressure from the x-ray source, increasing the velocity by a few percent at the outer radius.
Under the prescription of an outﬂowing gas Cloudy adjusts the density to conserve mass
ﬂux (deﬁned as the quantity ρ(r)r2 u(r), where ρ is the gas density, u is the velocity, and
53

6.0
(a)

(b)

10

5.5
16

30

4

5.0

20

5.0
8

4.5

24
8

ρH(R0) (log 1/cubic cm)

15

25

5.5

5

20

6.0

4.0

12

4.5

4

17.5

18.0
18.5
19.0
R0 (log cm)

19.5

3.5

17.5

18.0
18.5
19.0
R0 (log cm)

19.5

Figure 3.4 [O III]λ5007/Hβ ratio contours in Cloudy model parameter space of a solar
composition gas. The Cloudy models depend on three free parameters: inner radius of the
gas distribution (R0 ), hydrogen density at the inner radius (ρH (R0 )), and total gas mass.
Panel (a) shows the [O III]λ5007/Hβ ratio contours for 0.2 M⊙ gas mass. At this mass
the maximum ratio of 26 occurs at position indicated by the cross. A maximum ratio of
[O III]λ5007/Hβ =34 for all parameter space occurs for a gas mass of 100 M⊙ with contours
for the other two free parameters shown in panel (b).

54

r the radius) at each radius. The gas was irradiated by bremsstrahlung emission from a
central source at 106 K, and normalized to match the x-ray luminosity observed by Shih et
al. (2008); Maccarone et al. (2007). Parameter space was then explored along the dimensions
of the the inner radius (R0 ), hydrogen density at the inner radius (ρH (R0 )), and the total
mass of the outﬂow.
As an example of the resultant model space for one speciﬁc outﬂow mass ﬁgure 3.4a
displays the [O III]λ5007/Hβ emission line ratios for a model with a 0.2 M⊙ outﬂow as
a function of R0 and ρH (R0 ). Considering a 0.2 M⊙ outﬂow allows the examination of a
signiﬁcant limit, as for a solar composition gas, this is the minimum mass required to provide
the ∼ 10−4 M⊙ of oxygen required to produce the observed [O III] emission (Steele et al.,
2011). For this 0.2 M⊙ outﬂow a maximum emission line ratio of 26 occurs at R0 = 4.0×1018
cm and ρH (R0 ) = 1.0 × 105 gm cm−3 .
The emission line ratio rises gradually with increased outﬂow mass reaching a maximum
of [O III]λ5007/Hβ = 34 near 102 M⊙ as seen in ﬁgure 3.4b. The maximum emission
line ratio for a range of outﬂow masses is given in ﬁgure 3.5. The transition point in the
maximum emission line ratio-mass ratio seen at a mass of 1034 g occurs due a change in the
optimum geometric distribution of the emitting gas. For masses below the 1034 g threshold
the optimal arrangement involves locating the gas in a thin shell at a density which allows
maximum [O III]λ5007 emission across the shell. Above the threshold mass the optimal gas
conﬁguration is of a thick distribution, where the gas distribution thickness is greater than
the distance from the ionizing source and the [O III]λ5007 emission is generated over a broad
region located primarily in the half of the gas distribution most distant from the ionizing
source.
55

34

[O III]λ 5007/Hβ

32

30

28

26
32.5

33.0

33.5

34.0
34.5
Mass (log g)

35.0

35.5

Figure 3.5 Maximum [O III]λ5007/Hβ ratios by gas mass. The trend-line depicts the maximum [O III]λ5007/Hβ ratio from Cloudy models for any given total gas mass, allowing the
other free parameters (inner radius of the gas distribution and hydrogen density at the inner
radius) to vary as necessary. The maximum ratio occurs with a total gas mass of ∼ 100 M⊙ .

Table 3.2. [O III]λ5007/Hβ Ratio Limits
Conﬁdence (%)

Aperture (km s−1 )

Ratio, Upper Hβ Limit

Ratio, Lower Hβ Limit

68.3
90.0
95.0
99.0
68.3
90.0
95.0
99.0

3200
3200
3200
3200
600
600
600
600

52.84
39.96
35.70
29.55
41.05
33.04
30.17
25.78

∞
-163.9
-110.1
-67.06
165.4
7186.
-364.3
-119.3

56

3.4

DISCUSSION AND CONCLUSION

From the EW measurements presented in section 3.3.2 it is clear that any Hβ emission which
may be present in the observed RZ2109 spectrum is extremely weak. The Hβ emission is
suﬃciently weak to make interpretation of the resulting [O III]λ5007/Hβ line ratios diﬃcult.
If taken at face value these ratios are among the largest detected from any astrophysical
source (Sarzi et al., 2006; M´ndez et al., 2005). In table 3.3.3 the conﬁdence levels for the
e
[O III]λ5007/Hβ ratio are given based on the uncertainty in the Hβ measurement. Equivalent
width measurements below the level of the continuum (absorption features for example) are
indicated with negative equivalent width value. Therefore a negative line ratio limit in table
3.3.3 indicates that Hβ measurement is consistant with an absoprtion line at the specifed
conﬁdence level. From these limits the solar composition model cannot be ruled out above
the 95% level for the 3200 km s−1 aperture or the 90% level for the 600 km s−1 aperture,
however as considered below the masses and densities required to construct models that can
approach these limits make them exceedingly unlikely scenarios.
When the full velocity width covered by the [O III]λ5007 complex is considered along with
the radiative transfer modeling presented in section 3.3.3 it is clear that a solar composition
gas is insuﬃcient to produce the observed [O III]λ5007/Hβ ratios. The measured ratio using
3200 km s−1 apertures is nearly a factor of three larger than the maximum ratios produced
by the synthetic emission line models. At the 95% conﬁdence level for the 3200 km s−1
aperture and the 90% conﬁdence level for the 600 km s−1 aperture the uncertainty limits on
the [O III]λ5007/Hβ ratio approach the maximum synthetic [O III]λ5007/Hβ values. These
ratio limits are 35.7 at 3200 km s−1 , and 33.0 at 600 km s−1 compared to the synthetic
maximum of 34. As such it may be possible to produce the necessary [O III]λ5007/Hβ ratio
57

for either aperture given an emission line region model that falls in a very speciﬁc location
is physical parameter space. It should be noted, however, that the gas masses required
to produce the maximum synthetic [O III]λ5007/Hβ values are well above what might be
expected to be associated with an accreting black hole in a globular cluster. In order to
produce the maximum synthetic [O III]λ5007/Hβ ratios a total gas mass of order 100 M⊙ is
needed which would eliminate x-ray binaries, planetary nebulae, supernovae remnants, or any
other stellar scale objects as the source of the emitting gas. To reach a gas mass of that size
a signiﬁcant portion of the gas would necessarily be contributed by the cluster’s interstellar
medium. However the hydrogen densities involved in producing the maximum synthetic
[O III]λ5007/Hβ ratios are of order 104 –105 cm−3 , orders of magnitude above the expected
densities of a cluster interstellar medium. More typical gas masses produce maximum ratios
a factor of four times smaller than those observed in the 3200 km s−1 aperture and nearly
twice that of the lower limit of the 600 km s−1 measurement. The RZ2109 emission line
region must then necessarily be oxygen enriched relative to solar composition.

Steele et al. (2011) present the argument that the complex line [O III] velocity proﬁle
observed in the RZ2109 spectrum is consistent with emission originating from two discrete
gas structures. The two velocity apertures presented in this work do not directly correspond
to the two geometric components discussed by Steele et al. (2011) as emission from the
two components are superimposed in velocity space. From a comparison of the measured
[O III]λ5007/Hβ line ratios and synthetic line ratio considered here it seems most likely
that both the gas structures are comprised of oxygen enriched material. For the higher
velocity structure which Steele et al. (2011) describes as a bipolar conical outﬂow this is
consistant with the scenario of material being stripped from a CO white dwarf companion
58

to the black hole, and driven to the observed velocity as an accretion powered outﬂow. The
added constraint of being oxygen enriched, does not, however place obvious constraints on
the gas source for the lower velocity component.
As evidenced by the calculations summarized in ﬁgure 3.5, the geometry of the emission
region strongly inﬂuences the emission line ratios that it produces. As such the [O III]λ5007/Hβ
ratio alone is insuﬃcient to fully constrain the gas composition of the RZ2109 emission line
system, or to positively identify the particular stellar type of the x-ray binary’s donor star.
The WD donor star model is consistent with all the observations and calculations presented
above, however we are not yet able to fully rule out other late stage hydrogen depleted stellar types as the source for the observed outﬂow. With future observations of other spectral
bands, including measurements of UV carbon lines CIV 1548 & 1551 ˚ and CIII] 1907 &
A
1910 ˚, and more precise estimates of the emission line system’s gas mass it may yet be
A
possible to place tighter constraints on the emission line region’s composition and the x-ray
binary source system.

59

BIBLIOGRAPHY

60

BIBLIOGRAPHY

Kroupa, P. 2001, MNRAS, 322, 231
Ferland, G. J., Korista, K. T., Verner, D. A., et al. 1998, PASP, 110, 761
Gnedin, O. Y., Maccarone, T. J., Psaltis, D., & Zepf, S. E. 2009, ApJ, 705, L168
Ivanova, N. 2011, arXiv:1101.2864
Larsen, S. S. 2008, A&A, 477, L17
Le Borgne, D., Rocca-Volmerange, B., Prugniel, P., Lan¸on, A., Fioc, M., & Soubiran, C.
c
2004, A&A, 425, 881
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2007, Nature, 445, 183
M´ndez, R. H., Thomas, D., Saglia, R. P., et al. 2005, ApJ, 627, 767
e
Prugniel, P., Soubiran, C., Koleva, M., & Le Borgne, D. 2007, arXiv:astro-ph/0703658
Sarzi, M., Falc´n-Barroso, J., Davies, R. L., et al. 2006, MNRAS, 366, 1151
o
Shih, I. C., Maccarone, T. J., Kundu, A., & Zepf, S. E. 2008, MNRAS, 386, 2075
Stanghellini, L., Shaw, R. A., Balick, B., et al. 2003, ApJ, 596, 997
Steele, M. M., Zepf, S. E., Kundu, A., et al. 2011, ApJ, 739, 95
61

Zepf, S. E., Maccarone, T. J., Bergond, G., Kundu, A., Rhode, K. L., & Salzer, J. J. 2007,
ApJ, 669, L69
Zepf, S. E., et al. 2008, ApJ, 683, L139

62

Chapter 4
Variability of the [O III]λλ4959,5007
Emission Line Source in Black Hole
Host Globular Cluster RZ2109
Matthew M. Steele 1 , Stephen E. Zepf 1 ,Arunav Kundu 1 2 , Thomas J. Maccarone3 ,
Katherine L. Rhode 4 , and John J. Salzer 4

We examine the variability of the [O III]λλ4959,5007 emission line source in the NGC 4472
black hole hosting globular cluster RZ2109. Our continuing multi-facility monitoring program ﬁnds the strong emission line source had decreased 24 ± 2 percent from the 2007–2010
mean levels in 2011 and 40 ± 5 percent from the earlier mean in 2012. An analysis of the
1 Department of Physics & Astronomy, Michigan State University, East Lansing, MI

48824; e-mail: steele24@msu.edu
2 Eureka Scientiﬁc, 2452 Delmer Street Suite 100, Oakland, CA 94602-3017
3 School of Physics & Astronomy, University of Southampton, Southampton, Hampshire
S017 1BJ, UK
4 Department of Astronomy, Indiana University, Bloomington, IN 47405
63

variability of the emission line velocity proﬁle ﬁnds that the ﬂux ratio of higher velocity 1600
km s−1 component to the lower velocity 300 km s−1 component has decreased 30 percent
from 2009 to 2011, and the asymmetry between the red and blue wings of the proﬁle has
decreased 17 percent. We compare this variability to predictions of photoionized nova ejecta
models of the emission line region, and discuss its implications for an accretion powered
outﬂow from a CO WD-BH bianry model.

KEYWORDS: galaxies: individual (NGC 4472) — galaxies: star clusters: individual
(NGC 447 2) — globular clusters: general — X-rays: binaries — X-rays: galaxies: clusters

4.1

INTRODUCTION

The globular cluster RZ2109 is part of the part of the cluster system of Virgo Cluster
elliptical galaxy NGC 4472. Maccarone et al. (2007) originally identiﬁed RZ2109 as the ﬁrst
unambiguous detection of a black hole hosting globular cluster. The cluster is one of still
just a handful of globular clusters with identiﬁed black holes (Brassington et al., 2010; Shih
et al., 2010; Maccarone et al., 2011). The identiﬁcation of RZ2109’s black hole was made
using the variability of an x-ray source coincident with the cluster position. Maccarone et
al. (2007) argued that the 4 × 1039 erg s−1 luminosity could only be produced by a stellar
mass black hole accreting near the Eddington limit, or an ensemble of similarly accreting
neutron stars. Since the x-ray source was observed to vary by a factor of 7 over a period of
hours, the ensemble of neutron stars was ruled out as a possible source as such a collection
would not vary in unison.
Subsequent observations of the optical spectrum of RZ2109 have revealed the system
64

also plays host to an unusual [O III] emission line system. Observations published in Zepf
et al. (2007) and Zepf et al. (2008) note the presence of broad [O III]λλ4959,5007 lines with
widths of ∼ 2000 km s−1 and a luminosity of 1.4 × 1037 erg s−1 , and remarkably no other
detectable emission lines. Zepf et al. (2008) argued that the source of the emission lines was
a wind driven across the cluster and photoionized by the accreting black hole x-ray source.
Subsequent work by Steele et al. (2011, hereafter S11) and Steele et al. (2012, hereafter S12)
demonstrated that measurements of the [O III]λλ4959,5007 emission line structure appeared
to be constant over a 467 day period from December 2007 to March 2009 suggesting that
upper limits on the Hβ line measurements rule out the observed emission lines originating
from a solar composition gas. These observations lead to the suggestion that a CO white
dwarf-black hole binary is producing a photoionized accretion driven wind which is the source
of the observed [O III] emission.
Ripamonti & Mapelli (2012) propose a nova ejecta photoionized by the RZ2109 x-ray
source as an alternative model for the observed emission line system. In their work the
authors examine numerous possible conﬁgurations of the nova shell and construct predictions
for the luminosity evolution of a number of emission lines for a selection of possible models
produced by the expansion of the nova ejecta. As part of an ongoing study of the RZ2109 xray source Maccarone et al. (2010) note that the observed x-ray ﬂux has faded by a factor of
at least ten in observations obtained in 2010 compared to the earlier observations from 2004.
Under either the CO WD-BH binary outﬂow model or the nova shell model, such a drop in
ionizing photons should lead to a measurable decrease in the system’s [O III]λλ4959,5007
production.
In this study we consider three observations of RZ2109’s optical spectrum subsequent to
65

the S11 study. We analyze the [O III]λ5007 equivalent width variability over the combined
set of data points, and examine the emission line’s velocity proﬁle in comparison with previous observations. We then consider the implications of this variability for models of the
[O III]λλ4959,5007 emission line region.

4.2

OBSERVATIONS AND DATA REDUCTION

This work considers the ongoing multi-facility spectroscopic monitoring program of the globular cluster RZ2109 emission line system. As part of this program observations were made
using three facilities over a two year period from December 2009 to March 2012.

4.2.1

STIS

Optical spectra of RZ2109 were obtained from the Space Telescope Imaging Spectrograph
(STIS) as part of HST Program 11703. The observation program, using the G430L grating
and the 50x0.2 slit, was conducted over four dates (UT 2009 December 29, 2010 January
1, March 3, and April 11) and totaling 11 hours of integration time. The resulting spectra,
initially described by Peacock et al. (2012) were reduced using the standard STSDAS software
package. All the two dimensional exposures collected on a given date were co-added and
extracted to a one dimensional form. The spectra from each of the four dates were then
co-added and continuum normalized. As an early program using STIS in HST Cycle 17
after the instrument’s reactivation following Service Mission 4, the complications posed by
decreased charge transfer eﬃciency and other issues were not fully appreciated prior to the
observation program’s execution, producing signal-to-noise below the expected values.
66

4.2.2

Gemini

Gemini Program GS-2011A-Q-41 gathered optical spectra using the Gemini South facility
with the Gemini Multi-Object Spectrograph (GMOS) (Hook et al., 2004). The program
was executed as a queue observation on the nights of UT 2011 March 2,4, and 5 for a total
11.5 hr integration time. GMOS was set up using the B1200 grating and 1.0 arcsec slit
width working in longslit mode and realized a spectral resolution of R∼ 2600. The data
were reduced using the standard routines from the Gemini IRAF package. One dimensional
spectra were extracted from individual exposures, co-added, and continuum normalized.

4.2.3

SOAR

Observations were conducted using the Goodman High Throughput Spectrograph (Clemens
et al., 2004) on the 4.1m Southern Astrophysical Research Telescope (SOAR) on the nights of
UT 2012 March 14, 24. The instrumental setup included use of the KOSI 600 line/mm grating
and an 0.84 arcsec width slit. A total integration time over the two nights of 11.5 hours
produced a measured resolution of R ∼ 1500. The individual integrations were processed
with the standard IRAF NOAO Longslit tools, co-added by night, and extracted to one
dimensional spectra. The individual night spectra were then co-added and normalized to
the continuum.

4.2.4

Stellar Component Model

The observed spectrum of RZ2109 is a superposition of two components; a stellar component
contributed by the cluster member stars, and an emission line system component. In order
to study the emission line system of interest in this paper it is necessary to identify and
67

remove any contributions from the stellar component from the wavelength range over which
emission lines are found. This is done by constructing a synthetic stellar population model
based on a ﬁt of synthetic spectra to the regions of the stellar continuum where emission
lines are not expected to be found.
For construction of synthetic spectra the Pegas´-HR code of Le Borgne et al. (2004) was
e
employed along with the Elodie 3.1 (Prugniel et al., 2007) stellar library. A single epoch of
star formation was adopted along with a Kroupa IMF (Kroupa, 2001). For details of the
ﬁtting procedure see S11 and S12, for consistency we adopt the stellar parameters found in
these works (age = 13.25 Gyr, [Fe/H] = -1.0).

4.3

VARIABILITY

In examining the variability of [O III] emission from RZ2109 the data presented above are
combined with published data from S11. Since the [O III]λλ4959,5007 lines are produced
by the same excitation level of [O III] their line proﬁles should be identical and the ratio
of the lines determined by their relative de-excitation probabilities. Except where noted
the equivalent width and velocity proﬁle data are derived using the [O III]λ5007 line, the
stronger of the two lines.

4.3.1

Equivalent Width Variability

The equivalent width of the [O III]λ5007 line was measured by direct integration of the
velocity proﬁle of the entire [O III]λλ4959,5007 doublet and scaled to [O III]λ5007 using the
expected [O III]λ5007 / [O III]λ4959 ratio. To remove any contribution of the stellar component, an identical measurement is performed on the best-ﬁt synthetic stellar population
68

Table 4.1. [O III]λ5007 Equivalent Widths
Observation

EW (˚)
A

Date

Keck
WHT
SOAR
GMOS
STIS
GMOS
SOAR

30.9 ± 0.7
42.6 ± 2.5
26.8 ± 8.0
34.0 ± 0.4
38.5 ± 9.1
25.5 ± 0.4
20.1 ± 1.7

2007
2008
2009
2009
2010
2011
2012

∆Date (days)
0.0
20.0
434.1
467.0
786.0
1232.0
1549.0

spectrum derived in section 4.2.4 and subtracted for the data measurement. The results
of the equivalent width measurements for the 2010 STIS, 2011 Gemini, and 2012 SOAR
observations are summarized along with previous observations published S11 in ﬁgure 4.1.
Combined these observations represent over 4 years of equivalent width monitoring data.
Uncertainty estimates for the 2011 Gemini and 2012 SOAR observation were produced
by using an estimate of systematic uncertainties (instrumental noise, cosmic ray subtraction
errors, etc.) derived using root-mean-squared deviations from two regions of the continuum,
one redward of the [O III]λλ4959,5007 complex and one blueward, where stellar absorption
features were minimal and photon statistics. The two contributions were added in quadrature
to produce the uncertainty estimates given in table 4.1. For the 2010 STIS observation the
cited uncertainty was derived using the error values produced by the STSDAS reduction
pipeline over the measured wavelength aperture.
In the previous work on the [O III]λ5007 equivalent width variability, S11 argue that
from the 2007 Keck observation to the 2009 Gemini observation the data are consistant with
constant [O III]λ5007 ﬂux. With the inclusion of the STIS data point the [O III]λ5007 data
remain consistent with constant ﬂux through day 786 (in 2010). The uncertainty weighted
69

Continuum Normalized Flux

3.0

2.5

2.0

1.5

1.0

4900

4950

5000
Wavelength (˚)
A

5050

5100

Figure 4.1 RZ2109 [O III]λλ4959,5007 Emission Line Complex. The black line displays the
2009 Gemini observation, the red line gives the 2011 Gemini observation. Both spectra were
collected using the same instrumental setup and similar integration times. Note that the
[O III] ﬂux is lower at all wavelengths in the 2011 observation.

70

mean for the 2007-2010 equivalent widths is 33.3 ± 0.3 ˚. The 2011 Gemini observation
A
provides the ﬁrst strong departure from this mean, deviating by 24±2 percent from the mean.
Figure 4.1 provides a direct comparison between observations from the Gemini Observatory
using the same instrument and setup. The equivalent width of the 2011 observation has
decreased by 25 ± 2 percent from the 2009 observation. The 2012 Soar observation indicates
that the [O III] ﬂux has continued to decline, with a measured equivalent width 40 ± 5
percent below the 2007-2010 uncertainty weighted mean.

4.3.2

Velocity Proﬁle Variability

The velocity proﬁle of the [O III]λλ4959,5007 emission lines contains a wealth of information
regarding the velocity and geometric structure of the emission line region. In Steele et al.
(2011) the velocity proﬁle of the [O III]λλ4959,5007 was used to construct a simple geometric
model of the emission region comprised of two components; a broader 3200 km s−1 wide
ﬂat topped velocity structure which was well described by bipolar conical outﬂows, and a
narrower 600 km s−1 Gaussian velocity structure which did not include a unique signature of
a speciﬁc geometric structure. The simple geometric model presented in Steele et al. (2011)
derives its velocity proﬁle from projection of the outﬂow’s velocity relative to the line of
sight and contains no constraint on spatial scale or the radial gas density distribution. The
presence of variability in the [O III] velocity proﬁle could then be used to constrain both
these parameters.Since the two components in the S11 model are not explicitly continuous
they may vary on diﬀerent timescales, either because of separate scale lengths or diﬀerent
distances from the ionizing x-ray source.
To test the variability of the [O III]λλ4959,5007 emission line velocity proﬁle we examine
71

[O III]λ5007 EW Normalized Flux

0.06

0.04

0.02

0.00

−0.02
4900

4950

5000
Wavelength (˚)
A

5050

5100

Figure 4.2 Scaled [O III]λλ4959,5007 Emission Line Complex. The spectra displayed in this
plot have been normalized to their respective [O III]λ5007 equivalent width measurements to
allow a comparison of the relative velocity proﬁles of the [O III]λλ4959,5007 complex. The
black line gives the 2009 Gemini observation, the red line displays the 2011 Gemini data.

72

the diﬀerences between the Gemini observations from 2009 and 2011. The velocity proﬁles
are displayed in ﬁgure 4.2 scaled relative to their respective [O III]λ5007 equivalent widths
with the mean continuum level set to zero. In this conﬁguration the total area under the
[O III]λ5007 line proﬁle is set to unity, so the plotted value at any given wavelength displays
the portion of the total emission line ﬂux represented by that wavelength and is not indicative of the absolute observed ﬂux. Upon inspection two deviations from the 2009 Gemini
observation become apparent in the 2011 data; ﬁrst the narrower velocity component represents more of the total ﬂux relative to the broad component in the 2011 data, and second the
2011 data display a noticeable decrement in the broad component redward of the narrower
component centered near 5038 ˚ . These two deviations appear in both the [O III]λ4959 and
A
[O III]λ5007 lines, suggesting they are real.

As a test of signiﬁcance of of these deviations we bin each of the two equivalent width
normalized observations in 2 ˚ bins, roughly corresponding to the 1.9 ˚ FWHM of the
A
A
Gemini resolution elements, and examine diﬀerence between them. An estimate of the
expected one σ rms uncertainty for a bin is constructed using two 100 ˚ regions where no
A
variability is expected, one one the red side of the [O III]λλ4959,5007 emission lines and one
on the blue. In ﬁgure 4.3 the resulting diﬀerences are given as multiples of the expected 1 σ
RMS deviations. For the 64 bin comparisons there are a total of 26 deviations above the 1 σ
level, 7 above the 2 σ, and 4 above the 3 σ. For comparison ﬂuctuations consistent with RMS
noise should produce on average 20 bins above 1 σ, 3 above 2 σ, and less than one above 3
σ. The higher relative ﬂux in the narrower velocity component of the 2011 observation as
well as the decrement near 5038 ˚ are suﬃcient to account for the diﬀerences from expected
A
bin RMS statistics. The bins with the greatest deviations are signiﬁcant at the 6.4 σ level
73

8
6

Deviations (σ)

4
2
0
−2
−4
−6
−8
4960

4980

5000

5020
5040
˚)
Wavelength (A

5060

5080

Figure 4.3 Statistical Signiﬁcance of [O III]λλ4959,5007 proﬁle variability. The black line give
the statistical signiﬁcance of the diﬀerence between the 2009 and 2011 Gemini observations
of the RZ2109 [O III]λλ4959,5007 line proﬁle in 5 ˚ bins. For reference the 2009 observation
A
is given as a blue dotted line, the 2011 observation as a green dotted line.

74

Normalized Flux

2.5

2.0

1.5

1.0
−2000

−1500

−1000

0
500
−500
Velocity (km/s)

1000

1500

2000

Figure 4.4 Geometric Model Velocity Proﬁle. The Steele et al. (2011) geometric line proﬁle
model is ﬁt to the 2011 Gemini [O III]λ5007 velocity proﬁle (black line). All geometric
parameters are the same as found by Steele et al. (2011) for the 2009 Gemini data. The
bipolar conical outﬂow (blue line) has an opening angle of 70 degrees, inclination of 65
degrees, and velocity of 1600 km s−1 . The receding/approaching outﬂow ratio is 1.15. The
Gaussian component (green line)has a FWHM of 310 km s−1 . The two components are
co-added to produce the modeled velocity proﬁle displayed using a red line.

for the narrower velocity component, and the 5.2 σ level for the 5038 ˚ feature.
A
The geometric model presented in S11 allows for another means of examining the variability in the velocity structure. By adopting the geometric model we may ﬁt the 2011
observation with two free parameters, the lower velocity Gaussian component/higher velocity bipolar conical component normalization and the approaching cone/receding cone
normalization. The 2011 Gemini model ﬁt is displayed in ﬁgure 4.4, and shows that after
75

adjusting these two normalizations the same geometric model remains a good ﬁt everywhere
except around the 5038 ˚ ﬂux decrement. In this geometric model ﬁt 75 percent of the total
A
[O III]λ5007 ﬂux originates from the bipolar conical outﬂow, with the remaining portion
produced by the Gaussian component. For comparison the bipolar conical outﬂow/Gaussian
component ratio in the 2009 observations was 81/19 (S11). The receding/approaching outﬂow ﬂux ratio in the 2011 data is 1.15 compared to 1.39 in 2009. It should be noted that
the relative decrease in ﬂux in the red wing of the broader component which corresponds to
the receding conical outﬂow is at the 1 to 2 σ level for most 2 ˚ bins in the above analysis.
A

If the model-dependent ratios of the bipolar conical outﬂow to the Gaussian component
are used to calculate the contribution of each component to the equivalent width measurement, we ﬁnd that for the 2009 Gemini observation the bipolar conical component contributes
27.5±0.3 ˚ and the Gaussian 6.5±0.1˚. In the 2011 Gemini observations the bipolar conical
A
A
component contributes 19.1 ± 0.3 and the Gaussian 6.4 ± 0.1 ˚. By this analysis the bipolar
A
conical component has decreased in ﬂux by 31 ± 2 percent between 2009 and 2011 and the
Gaussian component is consistent with remaining unchanged. Since S11 demonstrate that
the line’s velocity proﬁle was constant from the 2007 Keck observation through the 2009
Gemini observation, it seems reasonable to conclude that the lower velocity Gaussian component has maintained a constant ﬂux from 2007 through at least 2011. Unfortunately the
velocity proﬁle of the 2012 SOAR data point is of insuﬃcient signal to noise to be reliably
ﬁtted to the geometric model and so the relative contributions of the two components cannot
be determined. If the assumption is made that the lower velocity Gaussian component remains unchanged through the 2012 SOAR observation and all variability is attributed to the
contributions from the higher velocity bipolar conical outﬂow, the bipolar conical outﬂow
76

varies by 59 ± 4 percent compared to the 2007–2010 uncertainty weighted mean.

4.3.3

VARIABILITY ANALYSIS

There are three major time scales which play a role in the observed [O III]λλ4959,5007 equivalent widths and velocity proﬁles; the ionizing ﬂux variability timescale, the light crossing
timescale, and the recombination timescale. In systems where the light crossing timescale
is suﬃciently large in comparison with the other two timescales it is possible use the delay
between the light curve of the ionizing source and that of the [O III]λλ4959,5007 emission to
measure the spatial size of the emission line region, similar to reverberation mapping of AGN
(Blandford & McKee, 1982; Peterson, 1993) or light echo measurement of media surrounding
novae and supernovae (Tylenda, 2004; Rest et al., 2005). The emission line velocity proﬁle
variability provides a simple check on the comparison of the light travel timescale. If the
light crossing timescale is suﬃciently long in comparison to the recombination timescale the
diﬀerence in light travel distances to the observer across an extend source will also manifest
itself in the velocity proﬁle of the emission line region.
In ﬁgure 4.6 the expected velocity proﬁle variability is displayed for the bipolar conical
outﬂow portion of the S11 geometric model. In this calculation the emission line region is
illuminated by a central ionizing source which varies in accordance with the observed x-ray
curve from Maccarone et al. (2010) displayed in ﬁgure 4.5. The four line proﬁles displayed
show the evolution of the emission line as the transition from the high x-ray ﬂux ﬂux regime
(∼ 4.5 × 1039 ergs s−1 ) to the low x-ray ﬂux regime (∼ 1.0 × 1039 ergs s−1 ) propagates across
the emission line region. The diﬀerence in light travel time between the approaching and
receding outﬂows and the observer results in the blue wing of the emission line proﬁle fading
77

X-Ray Flux (10ˆ39 ergs/s)

6
5
4
3
2
1
0

[O III]λ5007 EW (˚)
A

50
45
40
35
30
25
20
15

1995

2000
2005
Observation Date

2010

Figure 4.5 X-ray and [O III]λ5007 light curves. The x-ray ﬂux measurements displayed in
the top panel are taken from Maccarone et al. (2010). The [O III]λ5007 equivalent width
measurements are published in Steele et al. (2011) and the present work.

78

before the red wing. This calculation assumes an instantaneous recombination timescale,
however the same red/blue asymmetry persists as long as the recombination timescale is
smaller that the light crossing time.
From the measurements of the [O III]λ5007 velocity proﬁle variability presented in section
4.3.2 there is no evidence for a blue to red progression of the emission line fading. There are
three basic categories of emission line region structures that would allow for the emission
line shape to remain constant while the total ﬂux varies: 1) models with a ionizing ﬂux
variability timescale much greater than either the light crossing time or the recombination
timescale, 2) models with a recombination time scale much larger than the light crossing
time or the ionizing ﬂux variability timescale, 3) models with the ionizing source external to
the emission line region with the emission line region positioned along the observer’s line of
sight to the ionizing source.
The ﬁrst category of model is almost trivial; a cloud with a light crossing time smaller
than the ionizing ﬂux time scale will see the same incident ionizing ﬂux in all regions of
the cloud, so its emission density varies uniformly and emission of approximately the same
proportionality will reach the observer from all locations simultaneously. As a result a study
of the variability of the emission line originating from this type of source would only provide
insight on the variability of the ionizing source. S11 calculate the minimum radius for the
emission line region necessary to produce the observed [O III]λ5007 ﬂux is a few ×10−3 pc,
with a light crossing time of several days. At the critical density for [O III]λ5007 production
of Ne = 7.0 × 105 at a temperature of 104 K (Osterbrock, 1989) that is necessary to achieve
this minimum size the recombination timescale is ∼ 1 day. The time resolution of the
Maccarone et al. (2010) x-ray light curve makes determining the precise timescale of x-ray
79

1.2

1.0

Relative Flux

0.8

0.6

0.4

0.2

0.0

5010

5020

5030
Wavelength (˚)
A

5040

5050

Figure 4.6 Velocity Proﬁle Variability. The bipolar conical outﬂow model line proﬁle is shown
under the inﬂuence of a variable photoionizing source. The Maccarone et al. (2010) light
curve has been approximated using a step function with high luminosity (4 × 1039 erg s−1 )
and low luminosity (1 × 1038 erg s−1 ) regimes. The diﬀerent proﬁles result from the diﬀerent
light travel time between regions of the emission line region and the observer. The solid red
line indicates the initial line proﬁle, the solid black line at time equal to 0.5 light crossing
time of the radius, the dashed line after 1.0 light crossing time, the dashed and dotted line
1.5 light crossing times, and the dotted line 2.0 light travel times.

80

variability diﬃcult, however it could conceivably be up to 5-6 yr. The major diﬃculty with
this category of models is the measurement of 5-6 pc of the RZ2109 emission line region by
Peacock et al. (2012) using spatially resolved STIS spectra. The light crossing time for this
measurement is at least a factor of 3 larger than the maximum allowable x-ray variability
timescale.
In the second category of models the recombination timescale dominates the emission line
variability timescale. For this type of model the observed emission line variability directly
traces the recombination timescale. For a crude estimate of the recombination rate in this
category of model we use the uncertainty weighted mean of the 2007-2010 [O III]λ5007 ﬂux
along with the 2011 Gemini data to estimate a decrease in ﬂux of ∼ 20 percent per year. If
this is all attributable to the recombination timescale, we expect a recombination timescale of
not more than 5 years. Again this is roughly a factor of three smaller than the light crossing
timescale implied by the Peacock et al. (2012) measurement, indicating the recombination
timescale is insuﬃciently large to avoid a change in the shape of the [O III] velocity proﬁle.
The third category is a special case, that of a particular conﬁguration where the emission
line region is positioned intermediate along the observers line of sight with an external
photoionizing source. This arrangement has the qualiﬁcation that the distance between the
ionizing source and the emission line region must be suﬃciently great that the ionizing ﬂux
can approximated by a plane wave with its wavefront perpendicular to the observer’s line of
sight (i.e. the distance to the ionizing source much larger than the extent of the emission line
region). Also the recombination timescale must either be spatially homogeneous across the
emission line region, or roughly symmetric about an axis deﬁned by the line of sight. In such
an arrangement the diﬀerence in light travel time scales to the approaching and receding
81

outﬂows is exactly counterbalanced by the diﬀerence in arrival times of the incident ionizing
ﬂux to the approaching and receding outﬂows. In this case the emission line variability could
track either the ionizing ﬂux timescale or the recombination time scale, whichever is longer.

Another possible tack is to suggest multiple structures within the emission line region. A
spatially small dense structure with emission line variability timescale dominated by the xray variability time scale, and an extended diﬀuse structure with its emission line variability
dominated by the recombination timescale. The Peacock et al. (2012) scale measurement is
a half light radius, indicating that it might be possible to have a smaller scale structure at
smaller radii which would be capable of varying on timescales much shorter than the 5-6 pc
light crossing time. There are, though, complications to this interpretation. First, Peacock
et al. (2012) ﬁnd no diﬀerence in the half-light radius of the full velocity width [O III] proﬁle
and the lower velocity component of the proﬁle, thus the natural two structures provided
by the S11 geometric model cannot directly correspond to this small radius structure/large
radius structure model. Second, the free electron density must not be a smooth function of
radius. There must be distinct high density and low density regions, otherwise there will exist
some region where the recombination timescale reaches parity with the light travel timescale
producing an observable variation in the emission line velocity proﬁle. An ionization front
could produce the necessary discontinuity in free electron density, however both high and low
density regimes must be able to produce suﬃent [O III] emission. Thus for this arrangement
to work there must be two discrete spatial/density structures each of which has two distinct
velocity components and the lower velocity component of each has a mechanism with allows
it a much longer variability timescale than the high velocity component; an arrangement
suﬃciently complicated to be problematic.
82

Table 4.2. Ripamonti & Mapelli (2012) Nova Model [O III]λ5007 Light Curves1
Model
A
B
C
D

Mar 2009
3.43
3.36
3.72
1.73

/
/
/
/

3.14
2.33
2.68
0.480

Dec 2012

Mar 2012,Interpolated

1.48 / 0.555
0.368 / 0.095
2.73 / 1.150
0.005 / 0

1.87 / 1.07
0.966 / 0.542
2.93 / 1.46
0.35 / 0.10

∆F2012−2009 (%)
-45
-71
-21
-80

/
/
/
/

-66
-77
-46
-80

1 erg s−1 cm−2

4.4

DISCUSSION

There exist two broad types of emission line regions from which the RZ2109 [O III] emission
lines could be produced; those where the gas which comprises the emission line region shares
the same originating system as the x-ray source such as outﬂows/accretion disks associated
with x-ray binaries, and incidental systems which share the host globular cluster with the
x-ray source but do not share a common progenitor such as supernova remnants or nova
ejecta. Zepf et al. (2008) and S11 consider the possibility of the later category but focus
primarily on the former. In the speciﬁc case of nova ejecta S11 cite decay timescales of nova
light curves as evidence against this type of object as a source of RZ2109 [O III], but do not
consider the external ionization of the nova.
In conducting a more thorough examination of the nova ejecta as emission line system
Ripamonti & Mapelli (2012) describe four nova ejecta models they select as representative
of the model parameter space and which satisfy the observational constraints of Zepf et al.
(2008) and S11. Their predictions of the [O III]λ5007 ﬂux observations are reproduced in
table 4.2 along with linear interpolations to the 2012 SOAR observation date. The two ﬂux
entries in each column of the ﬁgure represent the two possible shell expansion geometries
83

employed in their model calculations. The ﬁrst value represents a model where the nova shell
maintains a ﬁxed thickness as the shell expands radially, in the second the shell thickness
expands to conserve the ratio of the shell thickness to the shell radius.
When the observations of the RZ2109 [O III]λ5007 equivalent width variability in 4.3.1 are
compared to the predictions in 4.2 it is clear that nova models of class B and D are excluded at
a high level of conﬁdence. The 2012 SOAR observation of the [O III]λ5007 has faded by 41±5
percent compared to the 2009 Gemini observation, matching the predictions of the Ripamonti
& Mapelli (2012) model A with a constant shell thickness and model C with an expanding
shell thickness. If the assumption is made that the lower velocity Gaussian component’s ﬂux
level has remained constant and all variability is due to the bipolar conical component the
ﬂux diﬀerence between the 2009 Gemini observation and 2012 SOAR observation is 59 ± 5
percent, favoring the nova ejecta model A with an expanding shell thickness. It must be
noted here that Ripamonti & Mapelli (2012) adopt a constant ionizing ﬂux from the x-ray
source, assuming that the decrease in ﬂux observed by Maccarone et al. (2010) is attributable
to an increase in the column density along the observers line of sight. If the decreased x-ray
ﬂux is in fact produced by a drop in luminosity the predicted ﬂuxes summarized in table 4.2
are over predictions.
The velocity proﬁle variability presented in section 4.3.2 does not provide much more of
a constraint on the Ripamonti & Mapelli (2012) nova shell models. The models of interest
have densities of 2–8 × 104 cm−3 which produce recombination timescales in tens of days.
Similarly the ejecta radii of 6–9 × 1016 cm correspond with light travel timescales in tens of
days. Both these of timescales are signiﬁcantly smaller than the ionizing ﬂux timescale and
so will not display any velocity dependence in the shape of the emission line velocity proﬁle
84

as the ﬂux decreases.
Some additional comment is warranted on interpretation of the 2010 STIS equivalent
width measurement. Unfortunately due to technical diﬃculties with STIS around the time
of the observation the measurement uncertainties make it unclear if the measurement reﬂects
a consistency with the 2007–2010 mean or the beginning of the equivalent width decline
demonstrated in the 2011–2012 observations. The X-ray data roughly contemporaneous
with the observation as well as the remarkable consistency favor a value in the lower portion
of the uncertainty range. In any case the STIS data are valuable in demonstrating that the
RZ2109 [O III] ﬂux levels did not fall oﬀ rapidly to the 2011 Gemini levels following the 2009
Gemini observation. It is therefore most consistent with the observations that the decline
in ﬂux between the Gemini observations occurred at a rate comparable to that between the
2011 Gemini and 2012 SOAR observations.
Turning to the [O III]λ5007 velocity proﬁle variability there remain two features which
diﬀer between the 2009 and 2011 Gemini observations that warrant consideration; the ﬂux
decrement at 5038 ˚ and the change in the red wing/blue wing ratio of the higher velocity
A
component. Of the two features the ﬂux decrement at 5038 ˚ is the most perplexing,
A
as it breaks the symmetry of the emission line proﬁle and deviates from any geometric
model predictions. There are a number of possible sources for the ﬂux decrement; it could
be produced by an over-dense region that is responding to the decrease in ionizing ﬂux
faster than the rest of the emission line region, a case of local obscuration by a gray body
absorber, or possibly outﬂow along a particular projection angle from the source encountered
a shock front removing it from its previous velocity space. None of these possibilities is
particularly compelling or makes strong predictions for future observations. The change in
85

the red wing/blue wing ﬂux asymmetry is more promising, though the signiﬁcance of the
ﬂux decrement is weaker.
The origins of the red/blue ﬂux asymmetry have not previously been well understood,
though S11 considers several possible origins. One intriguing possibility is that the asymmetry results from a light travel time discrepancy between the approaching and receding
outﬂows. From the x-ray and [O III]λ5007 light curves found in ﬁgure 4.5 it is possible
that the x-ray ﬂux was already fading from the high ﬂux state during the 2007–2009 [O III]
observations. If the light travel time diﬀerence between the two outﬂows is suﬃciently large
the type of asymmetric response in the light curve displayed in ﬁgure 4.6 could be responsible for the initial red/blue ﬂux asymmetry. Advancing forward in time to the 2011 Gemini
observation, the receding outﬂow has begun to fade as a period of time has passed equivalent
to the light travel time between the outﬂows. Additionally this model provides a natural
explanation for the ﬂux decrement at 5038 ˚ as the velocity space and light travel time
A
position where a drop in ionizing x-ray ﬂux is propagating through the the emission line
region. We therefore initially conclude that the velocity proﬁle is not varying, modulo a
slight renormalization only because the the initial velocity proﬁle already displayed signs of
a variable ionizing ﬂux.
The variability of the RZ109 [O III]λ5007 equivalent width and velocity proﬁle are suﬃcient to further constrain alternate models of the emission line region as an accretion powered
outﬂow and as a nova ejecta shell photoionized by and external source, although inadequate to favor one over the other. Continued monitoring of both the x-ray source and
[O III]λλ4959,5007 emission will certainly distinguish between the two as they each make
very diﬀerent predictions for the observable [O III] ﬂux over the next 5 years.

86

BIBLIOGRAPHY

87

BIBLIOGRAPHY

Blandford, R. D., & McKee, C. F. 1982, ApJ, 255, 419
Brassington, N. J., et al. 2010, arXiv:1003.3236
Clemens, J. C., Crain, J. A., & Anderson, R. 2004, Proc. SPIE, 5492, 331
Hook, I. M., Jørgensen, I., Allington-Smith, J. R., Davies, R. L., Metcalfe, N., Murowinski,
R. G., & Crampton, D. 2004, PASP, 116, 425
Kroupa, P. 2001, MNRAS, 322, 231
Le Borgne, D., Rocca-Volmerange, B., Prugniel, P., Lan¸on, A., Fioc, M., & Soubiran, C.
c
2004, A&A, 425, 8 81
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2007, Nature, 445, 183
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2010, MNRAS, 409, L84
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2011, MNRAS, 410, 1655
Osterbrock, D. E. 1989, Astrophysics of Gaseous Nebulae and Active Galactic Nuclei (Mill
Valley, CA: Univ. Science Books)
Peacock, M. B., et al. 2012
Peterson, B. M. 1993, PASP, 105, 247
Prugniel, P., Soubiran, C., Koleva, M., & Le Borgne, D. 2007, arXiv:astro-ph/0703658
88

Rest, A., Suntzeﬀ, N. B., Olsen, K., et al. 2005, Nature, 438, 1132
Ripamonti, E., & Mapelli, M. 2012, MNRAS, 423, 1144
Shih, I. C., Kundu, A., Maccarone, T. J., Zepf, S. E., & Joseph, T. D. 2010, ApJ, 721, 323
Steele, M. M., Zepf, S. E., Kundu, A., et al. 2011, ApJ, 739, 95
Steele, M. M., Zepf, S. E., Kundu, A., et al. 2012, in prep
Tylenda, R. 2004, A&A, 414, 223
Zepf, S. E., Maccarone, T. J., Bergond, G., Kundu, A., Rhode, K. L., & Salzer, J. J. 2007,
ApJ, 669, L69
Zepf, S. E., et al. 2008, ApJ, 683, L139

89

Chapter 5
Summary and Conclusion
The globular cluster RZ2109 in the Virgo cluster galaxy NGC 4472 hosts a pair of unusual
radiative sources; a luminous and variable x-ray source, and a strong, broad [O III] emission
line system. The x-ray source is known to be associated with the ﬁrst unambiguously identiﬁed black hole in a globular (Maccarone et al., 2007). The class of astrophysical object from
which the observed emission lines originate has remained unidentiﬁed. Previous observations
have noted that emission line source displays very broad [O III] features with no other strong
emissions, properties which make the RZ2109 source without clear analog in either Galactic
or extragalactic systems (Zepf et al., 2007, 2008). The current study has sought to advance
the understanding of the nature of the [O III] emission line source in the black hole hosting
globular cluster RZ2109. To this end and as part of an ongoing multi-facility monitoring
program, optical spectroscopic observations were collected and over a period from 2007–
2012. To quantify the [O III] emission equivalent width measurements were performed on
all observations and the velocity proﬁle examined for all observations with suﬃcient signal
to noise. These observations were then analyzed to identify the geometric and kinematic
90

structure, and chemical composition of the emission line region. The resulting measurement
and models are used to describe the phenomenology of the emission line region and identify
a model for the origin of the gas structure and its relation to the globular cluster’s accreting
black hole x-ray source.

5.1

Summary of Observations

The RZ2109 [O III]λ5007 emission line ﬂux displayed two states during the observational
program; an initial constant ﬂux state followed by periods of decreasing ﬂux. From 2007–
2010 the [O III]λ5007 line displayed a nearly constant ﬂux. The uncertainty weighted mean
of the equivalent width over this time was 33.3 ± 0.3 ˚, with the two highest signal-to-noise
A
observations diﬀering by 9 ± 2 percent. Observations from 2011 displayed a decline of 25 ± 2
percent from the 2007–2010 level. By 2012 the [O III]λ5007 ﬂux had declined 40 ± 5 percent
from the constant ﬂux state.
Equivalent width measurements of the Hβ feature yield either weak emission (0.21±0.13 ˚
A
for a 600 km s−1 measurement aperture) or values consistent with non-detection (0.32 ± 0.32
˚ for a 3200 km s−1 aperture). The [O III]λ5007/Hβ emission line ratios produced are
A
among the highest of any known astrophysical system. For a 600 km s−1 aperture this ratio
is 61.6, for a 3200 km s−1 aperture the [O III]λ5007/Hβ ratio is 105.7.
The velocity proﬁle of the [O III]λ5007 line displays two distinct velocity structures;
a roughly Gaussian central peak with a width of 600 km s−1 , and 3200 km s−1 wide ﬂat
topped wings which are symmetric in their velocity extent and shape, but display diﬀerent
ﬂux levels. From 2007–2009 no statistically signiﬁcant variability in the velocity proﬁle was
observed. Observations from 2011 showed departures from the 2007–2009 velocity proﬁle
91

in three ways. First, the relative contribution from lower velocity component to the total
[O III]λ5007 ﬂux increased. Second, a ﬂux decrement appeared near 5038 ˚ in the red wing
A
of the velocity proﬁle. Third, the ﬂux ratio ratio of the red and blue wings changed, with the
red wing contributing less of the total ﬂux. Beyond these three changes, two of which may
be characterized as re-normalizations of the contributions of distinct structures, the shape
of velocity proﬁle remained unchanged.

5.2

Summary of Analysis

The stellar component of the RZ2109 spectrum was ﬁt with synthetic stellar population
model using a Kroupa IMF (Kroupa, 2001) and a single epoch of star formation. The
resulting best ﬁt model had an age of 13.25 ± 1.00 Gyr and [Fe/H] = −1.0 ± 0.1. When the
stellar component is subtracted from the observed system, the remaining [O III] emission
line system is well ﬁt by a simple geometric model comprised of two components; a bipolar
conical outﬂow and a Gaussian component. The bipolar conical outﬂow ﬁts the broad
wings of the observed [O III]λ5007 velocity proﬁle (best ﬁt parameters: outﬂow velocity =
1600+190 km s−1 , opening angle = 70 ± 8 degrees, and inclination angle = 65+3 degrees),
−90
−6
while the Gaussian component contributes to the lower velocity component (FWHM 310 km
s−1 ). When this geometric model is used to ﬁt the 2011 Gemini data the entire decline in
[O III]λ5007 ﬂux can be attributed to a decline in ﬂux of the bipolar conical outﬂow with the
lower velocity Gaussian component contributing equal ﬂux to the 2007–2009 observations.
It is impossible to reproduce the observed [O III]λ5007/Hβ emission line ratio with a
solar composition gas and reasonable mass and density limits using spectral synthesis code
Cloudy (Ferland et al., 1998) to model the RZ2109 emission line region. The gas composi92

tion in the RZ2109 emission line region must therefore be oxygen enriched relative to solar
composition. These constraints on composition, combined with the modeled emission line
region geometry and kinematics, are consistent with the RZ2109 [O III] emission originating
from a photoionized wind driven from CO WD-BH x-ray binary. If conﬁrmed by future
studies this would be the known WD-BH system.
Ripamonti & Mapelli (2012) have proposed an alternative model where the [O III] ﬂux
is produced in an expanding nova ejecta may be constrained by the [O III]λ5007 variability
data. Comparing nova model predictions to the observations constrains the possible parameter space to a class of models where the photoionizing source is internal to the ejecta shell.
Future observations will allow for distinguishing between the WD-BH binary model and the
nova ejecta model, as the nova ejecta model should continue to decline in [O III] ﬂux while
the [O III] emission line ﬂux for an outﬂow from a WD-BH x-ray binary will be set by the
variability of the incident x-ray ﬂux.

93

BIBLIOGRAPHY

94

BIBLIOGRAPHY

Kroupa, P. 2001, MNRAS, 322, 231
Ferland, G. J., Korista, K. T., Verner, D. A., et al. 1998, PASP, 110, 761
Maccarone, T. J., Kundu, A., Zepf, S. E., & Rhode, K. L. 2007, Nature, 445, 183
Ripamonti, E., & Mapelli, M. 2012, MNRAS, 423, 1144
Zepf, S. E., Maccarone, T. J., Bergond, G., Kundu, A., Rhode, K. L., & Salzer, J. J. 2007,
ApJ, 669, L69
Zepf, S. E., et al. 2008, ApJ, 683, L139

95

APPENDICIES

96

Appendix A
Emission Line Region Geometry and
Velocity Proﬁles

A.1

Motivation and Code Description

The velocity proﬁle of an emission line contains a wealth of information about the gas region
from which it originates; including the velocity, density, and geometric structures of the
emitting gas. Disentangling these elements can be a non-trivial task especially in the face
of the added complications of turbulent velocities, self absorption and scattering which can
mask the signatures of emission line regions structure. As such emission line proﬁles are
commonly fully analyzed in a few handful of speciﬁc object types, such as stars with strong
stellar winds like those with P Cygni proﬁles and Wolf-Rayet stars, or broad line regions in
active galactic nuclei.
The [O III]λλ4959,5007 emission line source in globular cluster RZ2109 has a number
of features which make its emission line complex a fruitful target of study. First the O
97

III emission lines observed, [O III]λ4959 and [O III]λ5007 , originate from the forbidden
transition between the 1 D2 and the 3 P1 and 3 P2 excited states of O III respectively. As
such several major hurdles in the analysis of the emission line region are removed, as selfabsorption and scattering are negligible for either emission line indicating that ﬂux of a
given velocity bin is directly proportional to the number of O III ions with the corresponding
velocity. A second major advantage of the RZ2109 [O III] complex is large ﬂux of the line.
With a ﬂux of order 1037 erg s−1 (Zepf et al. 2008) about 50 percent of the total ﬂux
from RZ2109 over the wavelengths with [O III]λ5007 emission originate in the emission line
system. This allows the emission line velocity proﬁle to be easily separated from the stellar
component of the total globular cluster spectrum, and provides reasonable signal-to-noise
over the full emission line proﬁle. Third, the broad nature of the line at over 3200 km s−1
means the line proﬁle contains multiple spectral resolution elements which may be ﬁtted by
a model. Finally, since [O III]λ4959 and [O III]λ5007 are produced by the same excited state
of [O III] they originate from the same gas distribution and will have identical line proﬁles
with their ﬂux ratio set by the de-excitation probabilities, allowing for a consistency check
of the velocity proﬁle in each observation.

The code used to analyze the RZ2109 [O III] emission line velocity was written to generically and ﬂexibly address the velocity proﬁles of forbidden emission lines. While the intent of
the tools is to model the speciﬁc object of interest for this work, it contains no speciﬁcations
exclusive to RZ2109 or the observational facilities and instruments used for this investigation. The code used to construct the modeled velocity proﬁles is fairly simple, consisting of
three major parts. The ﬁrst is a set of deﬁnitions which specify the geometry, kinematic,
and emissivity proﬁles. The second portion integrates the deﬁnitions producing a velocity
98

proﬁle relative to the observers line of sight. The third component simulates an observation
of the emission line, returning a spectrum with the ﬂux normalized to the maxim ﬂux bin
and a wavelength solution normalized about the central wavelength.
The velocity integration is fully relativistic, including relativistic redshifts and Doppler
boosting. Line broadening by natural line width or turbulent velocities are not included
in the calculations, but for lower velocity outﬂows where these mechanisms may contribute
signiﬁcantly to the emission line proﬁle the resulting model spectrum may be smoothed with
a kernel of the appropriate width to compensate. Since the intent of the code is working
with the velocity proﬁle of forbidden lines absorption mechanisms are not implemented. The
code itself is written in Fortran 95 with multi-threading. Python wrappers are used to call
speciﬁc geometric model calculations for ease of use in data modeling and analysis.
The following sections consider speciﬁc geometric models, describing the parameter space
used to deﬁne each model and demonstrating the range of velocity proﬁle which may be
produced by each geometric model. The geometric systems considered provide simpliﬁed descriptions of common astrophysics systems and include expanding spheres outﬂows, rotating
spheres, ﬂattened disk-like outﬂows, rotating disks, and bipolar conical outﬂows.

A.2

Spherical Outﬂow

An expanding sphere is among the simplest geometric structures an emission line region can
take. Figure A.1a displays a sphere with a constant velocity directed radially outward and
a uniform radial emissivity proﬁle. Since the the component of the velocity directed toward
the observer is the same for any radial column, all velocity bins contain equal ﬂux producing
a ﬂat topped velocity proﬁle. The only free parameter in such a model is the outﬂow velocity,
99

1.0

(b)

(a)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0

(c)

−1000 −500

0

500 1000

0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000
Velocity (km/s)

Figure A.1 Spherical Outﬂow. All models have a radially directed outﬂow velocity of 1000
km s−1 . In all panels the solid line indicates a inclination of the symmetry axis with respect
to the observer’s line of sight of 0 degrees, the dashed line 45 degrees, and the dotted line
90 degrees. Panel (a) displays an model with uniform emissivity, in panel (b) the emissivity
is proportional to the sine of the inclination above the mid-plane, for panel (c) an external
ionizing source produces an emissivity function that decreases linerarly with distance from
the incident surface of the ionizing ﬂux.

100

which sets the width of the velocity proﬁle. Introducing a radial dependence in the emissivity
function does not change the velocity proﬁle, as the ﬂux in any given velocity bin is simply
the integral of a radial column. Thus any radial dependence in the emissivity represents an
equal change in all velocity bins and the proﬁle retains the same shape.
If an angular dependence is introduced in the emissivity proﬁle the expanding sphere
is reduced to a single axis of symmetry and a second free parameter is introduced, the
inclination angle relative to the observer. Figure A.1b depicts the velocity proﬁle of a sphere
where the symmetry is broken by adopting an emissivity function proportional to the sine
of the inclination above the mid-plane. This emissivity is used to roughly estimate the
velocity proﬁle of a weakly conﬁned polar outﬂow. The black line represents the line of an
orientation where the observer’s line of sight is parallel to the polar symmetry axis. In this
conﬁguration the greatest emissivity is situated along the line of sight with velocities either
directed toward the observer or away from the observer, creating a double peaked velocity
proﬁle. The dotted line depicts an orientation where the polar axis is perpendicular to the
line of sight. In this conﬁguration the volume elements with peak emissivity have velocities
directed perpendicular to the observer’s line of sight and a rounded proﬁle is observed. The
dashed line shows the velocity proﬁle where the polar axis is inclined 45 degrees with respect
to the observer’s line of sight.
In ﬁgure A.1c the emissivity symmetry is broken by introducing a depth dependency in
the emissivity function. This emissivity proﬁle simulates a uniform sphere being ionized by an
external source, with the emissivity of each volume element being a function of the distance
through the sphere that was travelled by the ionizing photon. In the model presented here
the emissivity is reduced linearly such that a volume element 1 radius from the surface of
101

the incident ionizing ﬂux will produce 50 percent the emission as a volume element at the
surface. The solid line in panel c represents the ionizing source line of sight parallel to the
observer line of sight, in this set there is a clear symmetry breaking as the approaching side
of the sphere is preferentially ionized. The dotted line represents the ionizing source line
of sight perpendicular to the observer line of sight, this line appears symmetrical as equal
ionization occurs on the approaching and receding sides of the sphere. The dashed line
represents a 45 degree angle between the two lines of sight.

A.3

Rotating Sphere

A second common geometry in astrophysical objects is that of a rotating sphere. The rotating
sphere model as implemented has ﬁve free parameters; 1) an interior radius at which 2) a
velocity is deﬁned, 3) the inclination of the rotation axis relative to the line of sight, 4) the
power of the emissivity proﬁle α expressed as ρem (r) ∝ r−α , and 5) the power of the velocity
proﬁle α expressed as v(r) ∝ r−α .
Figure A.2a displays the velocity proﬁles sphere with a Keplerian rotation curve (β = 0.5)
and an inverse square density proﬁle (α = 2). The three velocity proﬁles indicate the
inclination angles of the rotating axis relative to the observer (solid line 0 degrees, dashed
45 degrees, dotted 90 degrees). The general velocity proﬁle for models of this type may
be described as a broad ﬂat topped core with extended wings. It should be noted for an
inclination angle of 0 degrees there is no velocity component along the line of sight and
the width of the line is determined solely by the resolution of the smoothing applied in
the simulated observation. Panel b displays the velocity proﬁles of spheres with Keplerian
rotation and constant density (α = 0). These proﬁles retain the same basic shape as those
102

1.0

(a)

(b)

(c)

(d)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000 −1000 −500
Velocity (km/s)

0

500

1000

Figure A.2 Rotating sphere. All models have velocities of 1000 km s−1 and radii of 10 percent
of the outer radius, except panel (d) where the velocity is 1000 km s−1 at the outer radius.
Solid lines indicate inclination of the symmetry axis relative to the observers line of sight of
angles of 0 degrees, dashed lines 45 degrees, and dotted lines 90 degrees. Figure (a) α = 2
(inverse square emissivity) β = 0.5 (Keplerian rotation), (b) α = 0 (constant emissivity)
β = 0.5, (c) α = 2 β = 0 (constant velocity), (d) α = 2 β = −1 (solid body rotation).

103

in panel a except with less prominent wings, since a smaller percentage of the total emission
originates from the inner radii.
Figures A.2 c and d have fewer analogs in astrophysical systems but are provide here
for reference. Panel c displays the velocity curve of a model with an inverse square density
proﬁle and a constant velocity proﬁle (β = 0). Panel d also employs a inverse square density,
but uses solid body rotation (β = −1).

A.4

Disk-Like Outﬂow

A disk-like outﬂow is observed in certain types of planetary nebulae or stars with high mass
loss stellar winds. This geometry is similar to a spherical outﬂow but with a truncation of
the outﬂow at a given altitude above the mid-plane. A disk-like outﬂow has three degrees
of freedom; 1) the velocity of the outﬂow, 2) the altitude angle above the plane at which
the outﬂow is truncated, and 3) the inclination of the symmetry axis. As with the spherical
outﬂow all radially dependent emissivity proﬁles produce an identical emission line velocity
proﬁle.
Example velocity proﬁles of disk-like outﬂows are depicted in ﬁgure A.3. As with previous
geometries the three velocity proﬁles in each panel represent diﬀerent inclinations of the
symmetry axis relative to the observer; the solid line 0 degrees, the dashed 45 degrees,
and the dotted 90 degrees. The two panels provide examples of diﬀerent truncation altitude
angles. Panel a has a truncation angle of 2.5 degrees, and panel b a truncation altitude angle
of 10 degrees. As the truncation altitude grows larger width of the features in the velocity
curve grow broader. This pattern, perhaps unsurprisingly, continues until an truncation
altitude of 90 degrees, when the disk-like outﬂow becomes a spherical outﬂow.
104

1.0

(a)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0

(b)

0.8
0.6
0.4
0.2
0.0
−1000

−500

0
Velocity (km/s)

500

1000

Figure A.3 Disk-like outﬂow. All models have a radially directed velocity of 1000 km s−1 .
Solid lines indicate an inclination of 0 degrees, dashed lines 45 degrees, and dotted lines 90
degrees. The models in panel (a) have a truncation angle of 2.5 degrees above the mid-plane,
panel (b) has a truncation angle of 10 degrees.

105

A.5

Rotating Disk

Rotating disks are one of the most common geometries for astrophysical systems. In the form
of accretion disks they are characteristic features of proto-stellar and young stellar system,
as well as accreting compact objects. Like the rotating sphere model, a rotating disk has
ﬁve free parameters;1) an interior radius at which 2) a velocity is deﬁned, 3) the inclination
of the rotation axis relative to the line of sight, 4) the power of the emissivity proﬁle α and
5) the power of the velocity proﬁle β. For these calculations the rotating disk is assumed to
be thin (with a radius much greater than its scale height).
The double peaked velocity proﬁle is an easily recognizable shape for an emission line.
So much so that a double peak is often taken as a “smoking gun” signature of rotation
without the acknowledgment that other gas geometries are capable of producing double
peaked proﬁles. It is important therefore to be cognizant of the distinctive features that
diﬀerent rotating disks from other double peaked velocity structures. Figure A.4 displays
rotating disk velocity proﬁles for a range of emissivity and velocity functions. Here again the
line type diﬀerentiates the inclination angle of the symmetry axis: the solid line 0 degrees,
the dashed 45 degrees, and the dotted 90 degrees.
Panel a depicts an inverse square emissivity and Keplerian rotation. In panel b the
models have a constant emissivity and Keplerian rotation. Note for either Keplerian rotation
model the peaks are no more than 40 percent of full height of the feature. The peak to full
line height ratio can be used to help distinguish disk rotation signatures from other double
peaked velocity proﬁles. Here again as with the rotating sphere the strength of the wings
diﬀerentiates the emissivity models. For reference constant velocity (panel c) and solid body
(panel d) rotation models are provided, both using inverse square emissivity functions.
106

1.0

(a)

(b)

(c)

(d)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000 −1000 −500
Velocity (km/s)

0

500

1000

Figure A.4 Rotating disk. All models have velocities of 1000 km s−1 and radii of 10 percent of
the outer radius. Solid lines indicate inclination angles of 0 degrees, dashed lines 45 degrees,
and dotted lines 90 degrees. Figure (a) α = 2 (inverse square emissivity) β = 0.5 (Keplerian
rotation), (b) α = 0 (constant emissivity) β = 0.5, (c) α = 2 β = 0 (constant velocity), (d)
α = 2 β = −1 (solid body rotation).

107

A.6

Bipolar Conical Outﬂows

Bipolar conical outﬂows make up a up a category of geometries observed in accreting compact
objects, some planetary nebula, and asymptotic giant branch stars. This geometric model
has four free parameters; 1) the opening angle of the cones (here deﬁned as the full angle
from one side of the cone to the other), 2) the inclination angle of the polar axis relative
to the observer, 3) the velocity of the outﬂow, and 4) a normalizing factor between the two
polar cones.
The velocity proﬁles produced by bipolar conical outﬂows are the most varied of any
geometry considered here. Distinct velocity proﬁle shapes include single peaks, two discreet
peaks, two blended peaks, and three peak proﬁles. They are also the geometry of greatest
interest for the study of RZ2109. As such a total of 16 models are presented representing four
opening angles and four inclination angles. In each ﬁgure panel a represents an inclination
angle of 0 degrees, panel b 30 degrees, panel c 60 degrees, and panel d 90 degrees. All models
have have outﬂow velocities of 1000 km s−1 and a normalization between the cones of unity.
In ﬁgure A.5 the opening angle is set at ten degrees. Panels a–c display two discreet
peeks and in panel d the the two have merged to a single peak. It is important to note that
for suﬃciently narrow cones the width of each peak is set by the opening angle of the cone
and instrumental resolution. In these models there is a degeneracy between the inclination
angle and the gas velocity in determining the position of the peaks. As such for narrow
emission lines produced by small opening angle outﬂows (such as jets) it is impossible to
determine the velocity of the outﬂow absent a constraint on the inclination angle relative to
the observer’s line of sight.
Figure A.6 displays models with an opening angle of 40 degrees. Here again the ﬁrst
108

1.0

(a)

(b)

(c)

(d)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000
−1000 −500
Velocity (km/s)

0

500 1000

Figure A.5 Bipolar conical outﬂow. The velocity proﬁle of models with opening angles of 10
degrees. Panel (a) displays an inclination angle of 0 degrees, panel (b) 30 degrees, panel (c)
60 degrees, and panel (d) 90 degrees. All models have outﬂow velocities of 1000 km s−1 .

109

1.0

(a)

(b)

(c)

(d)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000
−1000 −500
Velocity (km/s)

0

500 1000

Figure A.6 Bipolar conical outﬂow. The velocity proﬁle of models with opening angles of 40
degrees. All other model parameters are identical to those in ﬁgure A.5.

110

1.0

(a)

(b)

(c)

(d)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000
−1000 −500
Velocity (km/s)

0

500 1000

Figure A.7 Bipolar conical outﬂow. The velocity proﬁle of models with opening angles of 80
degrees. All other model parameters are identical to those in ﬁgure A.5.

three panels display two discreet peaks that merge to form one. With this wider opening
angle however the complete degeneracy between velocity and inclination is broken, as the
peaks widen and change proﬁle dependent on the inclination angle.
The 80 degree opening angle of the models in ﬁgure A.7 display a broad range of velocity
proﬁle shapes. Somewhat larger than the 70 opening angle used to model the RZ2109 outﬂow,
this set of models displays the same basic properties properties as the RZ2109 model. At
the inclination angle of 0 degrees depicted in panel a there are two distinct peaks. By an
inclination angle of thirty degrees the peaks are signiﬁcantly asymmetric as shown in panel
111

1.0

(a)

(b)

(c)

(d)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000
−1000 −500
Velocity (km/s)

0

500 1000

Figure A.8 Bipolar conical outﬂow. The velocity proﬁle of models with opening angles of
120 degrees.All other model parameters are identical to those in ﬁgure A.5.

b. In panel c, with an inclination angle of 60 degrees, the two peaks overlap forming a third
central peak. This conﬁguration where the projections of the two cones overlap in velocity
is a major contributor to reproducing the characteristic shape of the [O III] velocity proﬁle
of RZ2109. For the inclination angle of 90 degrees shown in panel d the velocity proﬁle of
the two cones have completely merged forming a single broad rounded central feature.
For the 120 degree opening angle of ﬁgure A.8 the same basic line types of velocity proﬁles
are present as those shown in A.7. The lines have, however, broadened noticeably for every
inclination angle and changed the details of their shapes in very measurable ways. For the
112

inclination angle of 90 degrees it is clear the the wide opening angle bipolar conical outﬂow
models are well on there way towards becoming spherical outﬂows.

A.7

Relativistic Eﬀects

It is worth noting that the majority of geometric models described in the previous sections
were had velocity proﬁles calculated for gas velocities of 1000 km s−1 . In the regime where
relativistic eﬀects are small examining only a single velocity is acceptable as changing the
velocity will only change the width of the velocity proﬁle, not the characteristic shape of the
proﬁle. At relativistic velocities however, the gas velocities can have a signiﬁcant eﬀect on
the shape of the observed velocity proﬁle.
In relativistic systems the observed ﬂux of a object (Fobs ) is related to the intrinsic ﬂux
produced (Fint ) by the function:
Fobs = δ k−α Fint
where
δ=

1
γ(1 − β)

Here γ is the Lorentz factor, β is the line of sight velocity component as a fraction of the speed
of light, α is the spectral index (here α = 0 for emission line sources), and k is a structural
parameter equal to 2 for continuous emission structures and 3 for discrete structures. This
Doppler boosting eﬀect ampliﬁes the ﬂux produced by approaching outﬂows and dampens
the observed ﬂux of receding ﬂows. Figure A.9 displays the eﬀects of Doppler boosting on a
bipolar conical outﬂow velocity proﬁle shape with the RZ2109 ﬁt parameters (opening angle
70 degrees, inclination 65 degrees). The solid line represents a velocity of 1 percent the speed
113

1.0

Relative Flux

0.8

0.6

0.4

0.2

0.0
−1.0

−0.5

0.0
Relative Velocity

0.5

1.0

Figure A.9 Relativistic eﬀects on a bipolar conical outﬂow. All velocity proﬁles represent a
model with an opening angle of 35 degrees, an inclination of 35 degrees and and inclination
of 65 degrees. The solid black line has a outﬂow velocity of 1 percent the speed of light, the
dashed line 10 percent and the dotted line 20 percent. The velocity axis has been normalized
to the outﬂow velocity of each model.

114

of light, the dashed line 10 percent, and the dotted line 20 percent. For ease of comparison
the velocity scales for all three proﬁles have been normalized to the outﬂow velocity. Beyond
the expected ampliﬁcation and dampening eﬀects of the approaching and receding sides note
that the central feature shifts red-ward with increasing velocity. Since the central feature is
produced by a superposition of the two cones’ velocity proﬁles, the changing shape of each
cone’s proﬁle produces a shift, with no change to the geometric parameters of the emission
line region.

A.8

Velocity Proﬁle Variability

In RZ2109 the photoionizing source displays several modes of variably on times scales of hours
to approaching a decade. Dependent on the light crossing time and recombination timescale
the velocity proﬁle of an emission line region photoionized by a variable source might be
expected to change shape in response. In ﬁgure A.10 a bipolar conical outﬂow model is
calculated under the inﬂuence of a variable photoionizing source. For this calculation the
light curve of the photoionizing source is approximated by a step function with a one order of
magnitude drop in luminosity. The emission line region is assumed to have a recombination
timescale much less than the light crossing time of the system, and the emissivity of each
volume element is taken to be proportional to the photoionizing ﬂux incident upon it. These
two assumptions are each special cases but helpful simplifying assumptions for illustrative
purposes. Each panel displays four light curves corresponding to diﬀerent times from when
the ionizing source transitions from the high luminosity mode to the low luminosity mode.
The red line gives the initial velocity proﬁle, the black line after 0.5 light crossing time, the
dashed line after 1.0 crossing time, the dash-dotted line after 1.5 crossing times, and the
115

1.0

(a)

(b)

(c)

(d)

0.8
0.6

Relative Flux

0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
−1000 −500

0

500 1000 −1000 −500
Velocity (km/s)

0

500

1000

Figure A.10 Variability of bipolar conical outﬂow. The velocity proﬁles of lines displaying
a one magnitude drop in emissivity, representing a corresponding decrease in the incident
ionizing ﬂux. All models have an opening angle of 35 degrees and a outﬂow velocity of 1000
km s−1 . In each panel the solid red line indicates the initial line proﬁle, the solid black
line at time equal to 0.5 light crossing time of the radius, the dashed line after 1.0 light
crossing time, the dashed and dotted line 1.5 light crossing times, and the dotted line 2.0
light travel times. Panel (a) depicts an inclination of 0, (b) 30 degrees, (c) 60 degrees, and
(d) 90 degrees.

116

dotted line after 2.0 crossing times. The four panels indicate four inclination angles; a is 0
degrees, b 30 degrees, c 60 degrees, and d 90 degrees. The ﬁrst feature to note is that in
any orientation the blue side of the velocity proﬁle responds ﬁrst as the light travel time to
the observer is shortest for the approaching side. Second, not all orientations display the
same function of integrated ﬂux with respect to time; the model in panel a responds more
quickly to the drop in ﬂux than the model in panel d, but it also takes longer to reach its
ﬁnal conﬁguration. Third, while the models are varying they produce unique velocity proﬁle
shapes; it is not simply a matter of the approaching cone decreasing in ﬂux followed by the
receding cone.

A.9

General Observations

It is instructive to note here that not all model and parameter permutations produce uniquely
identiﬁable velocity proﬁles. For example any rotating model when viewed along the rotation
axis produces a velocity proﬁle determined solely by the eﬀective resolution of the observation. Several other geometries are capable of producing line proﬁles which ﬁt into identiﬁable
categories: square topped line proﬁles (expanding spheres,rotating spheres), round topped
line proﬁles (expanding spheres, bipolar conical outﬂows, disk-like outﬂows), and double
horned proﬁles (rotating disks, disk-like outﬂows, bipolar conical outﬂows). Many of the
individual line proﬁles can be distinguished from similar proﬁles in the same category by
their speciﬁc shapes, however doing so may be limited by the constraints of instrumental
resolution, the observation signal-to-noise, and the detailed physics of the region.

117

Appendix B
Cloudy Modeling of [O III] Emission
Line Regions

B.1

Description of Cloudy

The modeling of emission line ratios from the RZ2109 emission line source contained in
this work was performed using the spectral synthesis code Cloudy of Ferland et al. (1998).
Speciﬁcally the following discussion pertains to version 08.01 of that code. In calculating
the synthetic stellar spectrum the user sets up a basic gas composition and structure, then
Cloudy performs a set of detailed radiative transfer calculations propagating a speciﬁed
initial incident ﬂux through the medium, updating the physical parameters of the medium
at each incremental radius from the illuminating source until certain stopping conditions
are reached. In order to account for back scatter and reprocessing of the incident ﬂux the
calculation is processed from the outer radius inward and iterated either to convergence or
a speciﬁed number of times.
118

Cloudy is as a well documented and mature program, for details on the full range of
options and the physics included the reader is referred to the three part “Hazy” documentation series by Ferland1 . In this section I discuss the speciﬁc usages and parameters employed
in the simulation of the RZ2109 emission lines and some of the useful outputs that help
constrain the simulated parameter space.

B.2

Relevant Parameters

Cloudy is a ﬂexible code with many usage modes and means of invocation. The most basic
of which is feeding a text ﬁle with the appropriate model parameters in from the command
line and collecting the output to a second text ﬁle.
The most common invocation takes the the following form:
$ cloudy.exe < input.in > output.out
where ’input.in’ is the input speciﬁcation ﬁle, and ’output.out’ is the name to which the
output parameters are written. As an an example of an input speciﬁcation ﬁle consider the
following example from the solar abundance model grid described in chapter 4.
set nend 3000
iterate 3
brems 6
luminosity 39.6 range 14.7059 to 588.2
stop mass 33.5
radius 17.5 vary
1 Available at www.nublado.org

119

grid 17.8 to 19.0 in 0.1 dex steps
hden 4.6 vary
grid 4.5 to 5.5 in 0.1 dex steps
//using default stopping condition of T=4000K
sphere
wind 300, mass=0 //velocity in km/s
turbulence 10 //velocity in km/s
cosmic rays, background
print line column
print line sort wavelength range 5006 to 5008
print last
This input ﬁle calculates a two dimension grid of models with varying inner radii and
initial hydrogen densities for a ﬁxed total gas mass. The total gas mass is a stopping condition
for this calculation, and care must be taken that that the code terminates using the total
gas mass condition and not any other default stopping condition, so the model calculations
reﬂect the intended values in parameter space.
The ﬁrst two lines specify parameters pertaining to how the calculation will be performed.
set nend 3000
iterate 3
The ﬁrst of these sets the maximum number of zones to be calculated. Zones are shells
with a thickness set such that the gas parameters can be taken as constant across the shell.
The thickness of shells is set adaptively and will vary with radius. If no number of zones is
speciﬁed the default is 1400. Since the calculation will stop if the maximum number of zones
120

is reach for the solar composition grid it is important to set the value to a suﬃciently high
number of zones. Unfortunately there is no way to determine the appropriate number of
zones a priori. The second line sets the number of iterations to be performed with the default
of 1. Two iterations are usually satisfactory, three avoids non-convergence for more extreme
models. There is an ’iterate to convergence’ command, however it can cause computational
times to become large for little gain in precision.
Then next two lines set the spectral energy distribution (SED) and luminosity of the
ionizing source.
brems 6
luminosity 39.6 range 14.7059 to 588.2
The ﬁrst of these speciﬁes a bremsstrahlung SED with a temperature of 106 K. The second
line gives a normalizing luminosity of 1039/6 erg s−1 integrated between 14.7059 and 588.2
Ryd. These values were taken from ﬁts of the Maccarone et al. (2007) observations.
The next set of commands sets the physical parameters used to specify the model grid.
stop mass 33.5
radius 17.5 vary
grid 17.8 to 19.0 in 0.1 dex steps
hden 4.6 vary
grid 4.5 to 5.5 in 0.1 dex steps
The ﬁrst line sets the stopping condition as a mass of 1033.5 g. The second line sets the
inner radius of the gas distribution at 1017.5 cm, however the ’vary’ statement indicates that
there will be multiple calculations performed using a range of radii, so the actual number
121

given is not relevant. Immediately following a ’vary’ command, a ’grid’ statement is given
specifying the range over which the parameter will be varied. In this case calculations will
be performed for inner radii from 1017.8 –1019.0 cm in increments of 0.1 dex. The following
two lines specify the range over which the hydrogen density will be set at the inner radius.
Here the range for the calculation is between 104.6 –4.6 cm−3 in 0.1 dex. It should be noted
that a stopping condition cannot be iterated. This means in order to populate the total gas
mass dimension of parameter space multiple instances of Cloudy need to be run, each with
there own input parameter speciﬁcation ﬁle listing a diﬀerent stop mass.
The next line in the input ﬁle is not read by Cloudy.

//using default stopping condition of T=4000K

The double backslash characters indicate a comment. This comment is included as a reminder
that the default temperature stopping condition is 4000 K. For this set of models the stopping
mass will be reached prior to the temperature in any zone dropping to 4000 K. However this
is not true for all of parameter space, for large total gas mass the stopping temperature will
have to be lowered using a command like:

stop temperature 1000

The next three commands specify details of the geometry and kinematics of the gas
distribution.

sphere
wind 300, mass=0 //velocity in km/s
turbulence 10 //velocity in km/s
122

The ’sphere’ command sets the geometry as spherical. Of course the geometry of the RZ2109
region is not expected to be spherical, however the diﬀerence between a spherical distribution
and an open distribution occurs primarily in the luminosity of the output lines, not in
the relative line ratios. The line luminosities can be scaled appropriately to account for a
diﬀerent covering factor post calculation. This however is not true if the gas is speciﬁed to
be “clumpy”, in which case an open geometry can change how the radiation is reprocessed.
The ’wind’ command sets an outﬂowing wind at 300 km s−1 at the inner radius. A velocity
of 300 km s−1 is chosen to match the lower velocity component of the RZ2109 [O III]λ5007
proﬁle. The diﬀerence between this velocity and the higher 1600 km s−1 of the higher
velocity component in the resulting computational model is minimal. This wind speciﬁcation
eﬀectively sets the gas density proﬁle of the model as in the presence of a gas velocity Cloudy
conserves the mass ﬂux (ρ(r)r2 v(r)) through each zone. In calculating the velocity at radii
beyond the inner radius Cloudy applies radiation pressure and gravitational forces from a
central object to modify the velocity at each zone. In order to eliminate gravitational forces
from changing the density proﬁle the central mass is set to 0. The ﬁnal line sets the turbulent
velocity of the gas to a typical value of 10 km s−1 .
The next line sets other sources of ionization and energy input, specifying that a background of galactic cosmic rays should be included.
cosmic rays, background
The ﬁnal set of commands speciﬁes the formatting of the output.
print line column
print line sort wavelength range 5006 to 5008
print last
123

The ’print’ statement controls how output is written to standard output and ’line’ indicates
the command refers to the emission line luminosities and ratios. Thus the ﬁrst statement
tells Cloudy that the emission line data should be printed in a column format. The next
line speciﬁes that the emission line data should be sorted by the line wavelength and to only
print lines between 5006–5008 ˚. Finally the last statement indicates that the data should
A
only be written for the last iteration of the calculation.

B.3

Solar Abundance Grid

The [O III]λ5007/Hβ emission line ratios for the full parameters space of the solar abundance grid have been included as ﬁgures B.1–B.24. The contours in each panel give the
[O III]λ5007/Hβ ratio while the axes show the initial radius R0 and hydrogen density at
the initial radius ρH (R0 ). Each panel represents a separate total gas mass. The maximum
[O III]λ5007/Hβ ratio for each total mass cross section of parameter space is indicated as a
cross.
It should be noted that the [O III]λ5007/Hβ ratio is a much stronger function of the
inner radius than of the other parameters over most of parameter space. This is because for
the majority of models in the solar abundance grid the thickness of the gas distribution is
small compared to the inner radius. Thus the location of the the inner radius is strongly
set by the distance from the ionizing source that maximizes the ionization fraction of O III,
while simultaneously minimizing the neutral H fraction.
In chapter 3 the discussion of the solar abundance grid include a consideration of the
maximum [O III]λ5007/Hβ as a function of total gas mass. This function is constructed
by taking the maximum [O III]λ5007/Hβ model (indicated by the cross) from each mass
124

interval to deﬁne the function. One interesting feature is the sudden elbow in the function
at a mass of 34.1 log(g). Viewing the [O III]λ5007/Hβ contours for the full three dimensional
parameter space make understanding this elbow considerably easier. In ﬁgures B.15 and B.16
the transition from M = 34.0 to M = 34.1 log(g) shows a sudden transition in the inner
radius and initial H densities of the maximum [O III]λ5007/Hβ model, suggests a transition
in the dominant gas structure for [O III]λ5007/Hβ maximization.
To investigate this transition further it is helpful to look at emissivity proﬁles of models.
During the course of a model execution Cloudy calculates numerous physical and radiation
parameters which are not output by default. To produce emissivity proﬁles it is necessary
to insert a line into the models input ﬁle. For example in an input model at a total mass of
34.0 log(g) the following line was inclined:
punch lines emissivity last "mass34.0.ems"
Cloudy ’punch’ uses punch statement to indicate information should be saved to a data ﬁle.
The ’lines’ speciﬁcation indicates that the information to be saved pertains to the emission
lines, while ’emissivity’ gives the data type of interest. As earlier the ’last’ term dictates
that only the ﬁnal iteration will be saved and ’mass34.0.ems’ is the name of the data ﬁle all
this will be saved to.
As part of the analysis performed for chapter 3 emissivity proﬁles were produced for the
maximum [O III]λ5007/Hβ models for each total gas mass. The plots of the [O III]λ5007/Hβ
emissivity ratios as a function of the cloud depth D, the distance from the inner radius R0
to the speciﬁed zone, are displayed in ﬁgures B.25–B.33. To investigate the source of the
elbow in the maximum [O III]λ5007/Hβ vs. total gas mass function and the reason for the
sudden transition in the optimum inner radius and H density noted above it is helpful to
125

examine how the emissivity curves behave at the transition points. The relevant ﬁgures are
the last panel of B.28 and the ﬁrst panel of B.29. At the transition point the emissivity
proﬁles of the maximum [O III]λ5007/Hβ models dramatically shift. In the lower mass
regime the [O III]λ5007/Hβ is optimized over a much greater percentage of the total cloud
depth. Doing so requires a large inner radius and low density to maximize [O III]λ5007/Hβ
over a relatively narrow shell. In the high mass regime the maximum [O III]λ5007/Hβ is
concentrated in a smaller region of the total cloudy depth, but higher ratios are produced
in that region. To get the proper gas densities at the speciﬁc radius to achieve this peak
production requires a thicker shell with higher H densities and a smaller inner radii.

B.4

CO Models

As a ﬁnal example it it useful to note how Cloudy handles gas composition. As a default
cloudy works with solar composition, so no explicit commands were necessary to speciﬁc
the gas composition. For the RZ2109 emission line source the evidence strongly suggests
that the emission line region is comprised of a hydrogen depleted gas. Cloudy has numerous
methods for and approaches to setting the model gas composition, with varying degrees
of granularity. The reader is referred to the Hazy documentation for a full description of
these tools. However while there are many tools to change either the relative or absolute
abundances of most elements, there is no way to set the hydrogen abundance to zero. In
fact many Cloudy speciﬁcations and calculations depend on the presence of hydrogen, not
least of which are the gas density settings and calculations. While one cannot scale down the
hydrogen abundance it is possible to scale up other abundances while scaling down hydrogen
density so that the models with identical gas mass densities to the solar models are achieved
126

with lower hydrogen fractions.
When investigating the possibility of a carbon and oxygen enriched gas for RZ2109 emission line region the following commands were used to increase CO abundances.
element scale factor oxygen 10
element scale factor carbon 10
These two commands modify the default solar abundances such that oxygen and carbon
are scaled to 10 times there solar abundances. No other abundances are changed by these
commands. Figures B.34 and B.35 provide an example calculation of three model grids at
a total gas mass of 32.6 log(g) for compositions of solar, 10 times solar CO, and 100 times
solar CO. At this gas mass the maximum [O III]λ5007/Hβ ratio increase by a factor of ∼ 3
–4.5 for every magnitude increase in CO.

127

log M =31.1 (log g)
5.2

3

5.2

log M =31.2 (log g)

18

6
12

5.0

18
4.8

4.8

18

12

21

15

4.6

21

6

15

9

9

3

4.4

3

4.4

18

12

4.6

6

ρH(R0) (log 1/cubic cm)

6

12

9

15

5.0

4.2
4.2
18.3 18.4 18.5 18.6 18.7 18.8 18.9 19.0 18.3 18.4 18.5 18.6 18.7 18.8 18.9 19.0
R0 (log cm)
R0 (log cm)
Figure B.1 Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.1–31.2

128

log M =31.4 (log g)

log M =31.3 (log g)
5.2

5.2

15

21

15
9

18
21

12

4.6

6

ρH(R0) (log 1/cubic cm)

18

18

4.8

4.8

4.6

6

12

6
5.0

12

5.0

3

4.4

4.4

4.2
4.2
18.3 18.4 18.5 18.6 18.7 18.8 18.9 19.0 18.3 18.4 18.5 18.6 18.7 18.8 18.9 19.0
R0 (log cm)
R0 (log cm)
Figure B.2 Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.3–31.4

129

log M =31.5 (log g)

log M =31.6 (log g)

5.4

5.4
18

3

18

18

5.0

4.8

6

6

21

9

5.0

21

5.2

9

4.8

4.6

3

4.6

21

ρH(R0) (log 1/cubic cm)

18

18

5.2

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.3 Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.5–31.6

130

log M =31.7 (log g)

15

5.2

21

9

12

5.0

21

3

3

6

4.8

4.8

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

6

ρH(R0) (log 1/cubic cm)

5.4

5.2

5.0

18

18

6

5.4

log M =31.8 (log g)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.4 Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.7–31.8

131

log M =31.9 (log g)

log M =32.0 (log g)
18

18

5.4

5.4

6
12

5.0

5.0

3

15

3

4.8

4.8

4.6

6

ρH(R0) (log 1/cubic cm)

18

21

5.2

9

21

9

5.2

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.5 Solar Abundance Grid [O III]λ5007/Hβ. log M = 31.9–32.0

132

log M =32.1 (log g)

log M =32.2 (log g)
5.4

5.2

5.2

20

20

16

21

12

18

5.0

5.0

6

21
4

12

4.8

3

ρH(R0) (log 1/cubic cm)

8

5.4

24

4.8

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.6 Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.1–32.2

133

log M =32.3 (log g)

log M =32.4 (log g)

5.4

5.4

12

5.0

4

4
24

4.8

4.8

8

ρH(R0) (log 1/cubic cm)

8

24

12

5.0

20

20

20

5.2

20

5.2

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.7 Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.3–32.4

134

log M =32.5 (log g)

log M =32.6 (log g)

8

5.0

4

8
12

24

12

5.0

4

16

ρH(R0) (log 1/cubic cm)

5.2

24

5.2

20

5.4

20

5.4

4.8

4.8

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.8 Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.5–32.6

135

log M =32.7 (log g)

12
4

24

16

8

5.0

24

5.2

8
20

5.0

4

ρH(R0) (log 1/cubic cm)

12

5.2

20

5.4

16

5.4

log M =32.8 (log g)

4.8

4.8

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.9 Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.7–32.8

136

log M =32.9 (log g)

log M =33.0 (log g)

20

24

8

5.2

24

8

12

12

5.0

5.0

4
4

ρH(R0) (log 1/cubic cm)

16

5.2

20

5.4

16

20

5.4

4.8

4.8

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.10 Solar Abundance Grid [O III]λ5007/Hβ. log M = 32.9–33.0

137

log M =33.1 (log g)

24

12

5.2

20

12

24
5.0

4.8

4.8

4.6

4

5.0

4

ρH(R0) (log 1/cubic cm)

8

8

5.2

20

5.4

16

20

5.4

log M =33.2 (log g)

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.11 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.1–33.2

138

log M =33.3 (log g)

20

20

12

20

5.4

16

5.4

log M =33.4 (log g)

24

16

24
4

5.0

5.0
4

ρH(R0) (log 1/cubic cm)

8

5.2

8

5.2

4.8

4.8

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.12 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.3–33.4

139

log M =33.5 (log g)
5.6

20

20

5.4

log M =33.6 (log g)

4

12

4

5.0

24

ρH(R0) (log 1/cubic cm)

4
12

24

8

5.2
5.0

20

5.4

12

12

20

16

5.2

4.8
4.8
4.6

4.6

4.4
4.4
17.8 18.0 18.2 18.4 18.6 18.8 19.0 17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)
R0 (log cm)
Figure B.13 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.5–33.6

140

log M =33.7 (log g)

log M =33.8 (log g)

5.6

5.6

4

5.2

24

4

24

4

ρH(R0) (log 1/cubic cm)

12

12

12

5.2

5.0

20

20

4

12

5.4

20

20

5.4

5.0

4.8

4.8

4.6

4.6

4.4
4.4
17.8 18.0 18.2 18.4 18.6 18.8 19.0 17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)
R0 (log cm)
Figure B.14 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.7–33.8

141

log M =33.9 (log g)
5.6

24

5.6

log M =34.0 (log g)

4

20

5.4
8

16

12

12

4.8

4

4.6

4.6

24

4.8

4

5.0

8

5.0

24

5.2

8

24

5.2

16

20

ρH(R0) (log 1/cubic cm)

5.4

4.4
4.4
17.8 18.0 18.2 18.4 18.6 18.8 19.0 17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)
R0 (log cm)
Figure B.15 Solar Abundance Grid [O III]λ5007/Hβ. log M = 33.9–34.0

142

log M =34.1 (log g)

log M =34.2 (log g)
5.6

5.4

5.4

4
16

20

28

8

5.2

8

24

5.2

24

5.0

24

5.0

8

20

4.8

4.8

24

12

16

ρH(R0) (log 1/cubic cm)

4

5.6

4

4
4.6

4.6

4.4
4.4
17.8 18.0 18.2 18.4 18.6 18.8 19.0 17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)
R0 (log cm)
Figure B.16 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.1–34.2

143

log M =34.3 (log g)

log M =34.4 (log g)

5.4

5.4

5.2

12
28

28

16
4.8

24

5.0

24

5.0

8

4.8

4

ρH(R0) (log 1/cubic cm)

20

5.2

8

4
24

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

20

4.6

8

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.17 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.3–34.4

144

log M =34.5 (log g)
5.4

8

24

5.4

log M =34.6 (log g)

20

5.2

4.8

4.8

4.6

4.6

20

5.0

30

25

5.0

15

ρH(R0) (log 1/cubic cm)

25

4

12

28

24

5.2

5

10

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.18 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.5–34.6

145

log M =34.8 (log g)

log M =34.7 (log g)

10

5.4

10

5.4

30

25

5.0

4.8

4.8

4.6

4.6

25

5.0

30

20

ρH(R0) (log 1/cubic cm)

15

25

15

25

5.2

5

5

20

20

5.2

10
5

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.19 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.7–34.8

146

log M =34.9 (log g)

5

15

30

5.0

25

ρH(R0) (log 1/cubic cm)

10

25

20

5.2

30

5.0

20

10

25

5.2

5.4

5

15

5.4

log M =35.0 (log g)

4.8

4.6

4.6

25

4.8

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.20 Solar Abundance Grid [O III]λ5007/Hβ. log M = 34.9–35.0

147

log M =35.1 (log g)
5.4

20

5.4

log M =35.2 (log g)

5

4.8

4.8

30

5.0

10

20

15

5.0

5

15

5.2

10

30

ρH(R0) (log 1/cubic cm)

25

25

5.2

25

4.6

4.6

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

Figure B.21 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.1–35.2

148

log M =35.3 (log g)
5.4

4.40

log M =35.4 (log g)

4.35

5

20

25

15

4.25

30

4.20

25

15

4.15

20

10

5.0
30

ρH(R0) (log 1/cubic cm)

4.30
5.2

4.8

18.0

18.4
R0 (log cm)

18.8

4.00

5

4.05

4.6

10

4.10

18.4

18.6
18.8
R0 (log cm)

Figure B.22 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.3–35.4

149

19.0

4.40

log M =35.5 (log g)

4.40

4.30
4.25

4.25

30

4.20

4.20

20

30

4.15

4.15

25

ρH(R0) (log 1/cubic cm)

4.30

30

4.35
30

4.35

log M =35.6 (log g)

4.00

18.4

18.6
18.8
R0 (log cm)

4.00
19.0

20

4.05

4.05

25

4.10

15

4.10

18.4

18.6
18.8
R0 (log cm)

Figure B.23 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.5–35.6

150

19.0

4.40

log M =35.7 (log g)

4.35

ρH(R0) (log 1/cubic cm)

25

30

4.30
4.25
4.20
4.15
30

25

4.10
4.05
4.00

18.4

18.6
18.8
R0 (log cm)

19.0

Figure B.24 Solar Abundance Grid [O III]λ5007/Hβ. log M = 35.7

151

1.45

log [O III]/Hβ Emissivity

1.40
1.35
1.30
log(M) =32.5

log(M) =32.6

log(M) =32.7

1.25

log(M) =32.8

1.45
1.40
1.35
1.30
1.25
11.4

12.0

12.6

13.2

11.4

12.0

log(D) log(cm)

Figure B.25 Solar Abundance Grid Emissivity. log M = 32.5–32.8

152

12.6

13.2

1.45

log [O III]/Hβ Emissivity

1.40
1.35
1.30
log(M) =32.9

log(M) =33.0

log(M) =33.1

log(M) =33.2

1.45
1.40
1.35
1.30
11.2

12.0

12.8

13.6 11.2

12.0

log(D) log(cm)

Figure B.26 Solar Abundance Grid Emissivity. log M = 32.9–33.2

153

12.8

13.6

1.45

log [O III]/Hβ Emissivity

1.40
1.35
1.30
log(M) =33.3

log(M) =33.4

log(M) =33.5

log(M) =33.6

1.45
1.40
1.35
1.30
12.0

12.8

13.6

12.0

log(D) log(cm)

Figure B.27 Solar Abundance Grid Emissivity. log M = 33.3–33.6

154

12.8

13.6

1.45

log [O III]/Hβ Emissivity

1.40
1.35
1.30
log(M) =33.7

log(M) =33.8

log(M) =33.9

log(M) =34.0

1.45
1.40
1.35
1.30
12.0

12.8

13.6

14.4

12.0

12.8

log(D) log(cm)

Figure B.28 Solar Abundance Grid Emissivity. log M = 33.7–34.0

155

13.6

14.4

1.6
1.5

log [O III]/Hβ Emissivity

1.4
1.3
1.2
1.1

log(M) =34.1

log(M) =34.2

log(M) =34.3

1.0
1.6

log(M) =34.4

1.5
1.4
1.3
1.2
1.1
1.0
10.5

12.0

13.5

15.0 10.5

12.0

log(D) log(cm)

Figure B.29 Solar Abundance Grid Emissivity. log M = 34.1–34.4

156

13.5

15.0

log [O III]/Hβ Emissivity

1.7
1.6
1.5
1.4
1.3
1.2
1.1
1.0
0.9
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1.0
0.9

log(M) =34.5

log(M) =34.7
12.0

log(M) =34.6

log(M) =34.8

13.5

15.0

12.0

log(D) log(cm)

Figure B.30 Solar Abundance Grid Emissivity. log M = 34.5–34.8

157

13.5

15.0

1.7
1.6
1.5
log [O III]/Hβ Emissivity

1.4
1.3
1.2
1.1

log(M) =34.9

log(M) =35.0

log(M) =35.1

1.0
1.7

log(M) =35.2

1.6
1.5
1.4
1.3
1.2
1.1
1.0

12

13

14

15

16

12

13

log(D) log(cm)

Figure B.31 Solar Abundance Grid Emissivity. log M = 34.9–35.2

158

14

15

16

1.6

log [O III]/Hβ Emissivity

1.4
1.2
1.0
log(M) =35.3

log(M) =35.4

log(M) =35.5

log(M) =35.6

0.8
1.6
1.4
1.2
1.0
0.8

13

14

15

16

13

14

log(D) log(cm)

Figure B.32 Solar Abundance Grid Emissivity. log M = 35.3–35.6

159

15

16

1.5

log [O III]/Hβ Emissivity

1.0
0.5
log(M) =35.7
0.0

13.5

15.0

16.5

log(D) log(cm)

Figure B.33 Solar Abundance Grid Emissivity. log M = 35.7

160

Solar CO, Line Ratio 25.7

5.4

5.4

30

16

5.2

45
60

45

12

75

5.0

5.0

15

24

16

30

4.8

20

4.8

15

60

4
8

20

5.2
12

ρH(R0) (log 1/cubic cm)

10 × Solar CO, Line Ratio 80.4

8
4

4.6

4.6
17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)
Figure B.34 CO Enriched Models.

161

17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)

100 × Solar CO, Line Ratio 357.
100

150

200

5.2

300

250

ρH(R0) (log 1/cubic cm)

5.4

5.0

350
4.8

4.6
17.8 18.0 18.2 18.4 18.6 18.8 19.0
R0 (log cm)
Figure B.35 CO Enriched Models.

162