arXiv:0801.0321v2 [astro-ph] 18 Jan 2008
A Self-Consistent NLTE-Spectra Synthesis Model of FeLoBAL
E. Baron1, Karen Leighly, Darko Jevremovic2, David Branch
University of Oklahoma, Homer L. Dodge Department of Physics and Astronomy, Norman,
OK 73019, USA
We present detailed radiative transfer spectral synthesis models for the Iron
Low Ionization Broad Absorption Line (FeLoBAL) active galactic nuclei (AGN)
FIRST J121442.3+280329 and ISO J005645.1-273816. Detailed NLTE spectral
synthesis with a spherically symmetric outflow reproduces the observed spectra
very well across a large wavelength range. While exact spherical symmetry is
probably not required, our model fits are of high quality and thus very large
covering fractions are strongly implied by our results. We constrain the kinetic
energy and mass in the ejecta and discuss their implications on the accretion
rate. Our results support the idea that FeLoBALs may be an evolutionary stage
in the development of more “ordinary” QSOs.
Subject headings: AGN: FIRST J121442.3+280329, ISO J005645.1−273816
Spectroscopic observations of quasars show that about 10–20% have broad absorption
troughs in their rest-frame UV spectra (see Trump et al. 2006, for example). These ab-
sorption lines are almost exclusively blueshifted from the rest wavelength of the associated
atomic transition, indicating the presence of an outflowing wind in our line of sight to the
1Computational Research Division, Lawrence Berkeley National Laboratory, MS 50F-1650, 1 Cyclotron
Rd, Berkeley, CA 94720-8139 USA
2Current Address: Astronomical Observatory, Volgina 7, 11160 Belgrade, Serbia and Montenegro
– 2 –
nucleus. The line-of-sight velocities range from zero to up to tens of thousands of kilometers
per second (e.g., Narayanan et al. 2004).
While understanding these outflows is of fundamental interest for understanding the
quasar central engine, it is also potentially important for understanding the role of quasars
in the Universe. The observation that the black hole mass is correlated with the velocity
dispersion of stars in the host galaxy bulge (e.g., Magorrian et al. 1998; Ferrarese & Merritt
2000; Gebhardt et al. 2000) indicates a co-evolution of the galaxy and its central black hole.
The close co-evolution implies there must be feedback between the quasar and the host
galaxy, even though the sphere of gravitational influence of the black hole is much smaller
than the galaxy. Energy arguments, however, show that is quite feasible that the black hole
can influence the galaxy; as discussed by Begelman (2003), the accretion energy of the black
hole easily exceeds the binding energy of the host galaxy’s bulge.
The nature of the feedback mechanism that carries the accretion energy to the galaxy
is not known. Since AGNs are observed to release matter and kinetic energy into their
environment via outflows, it is plausible that these outflows contribute to the feedback in
an important way. One of the difficulties in using quasar outflows in this context is that
they are sufficiently poorly understood that there are significant uncertainties in such basic
properties as the total mass outflow rate and the total kinetic energy.
What is the kinetic luminosity of the broad absorption line quasar winds? That turns
out to be very difficult to constrain. While the presence of the blueshifted absorption lines
unequivocally indicates the presence of high-velocity outflowing gas, the other fundamentally
important properties of the gas, including the density, column density, and covering fraction
are very difficult to constrain.
The density is difficult to constrain because the absorption lines are predominately
resonance transitions, and their strengths are not very sensitive to density. Without knowing
the density, the distance of the gas from the central engine cannot be constrained; the same
ionization state can be attained by dense gas close to the central engine, or rare gas far
from the central engine. Density estimates are possible when absorption lines are seen from
non-resonance transitions, but even then, they can differ enormously. For example, de Kool
et al. (2001) analyzed metastable Fe II absorption lines in FBQS 0840+3633, and inferred
a electron density < 1000 − 3000cm−3and a distance from the central engine of several
hundred pc. In contrast, Eracleous et al. (2003) analyze the metastable Fe II absorption
in Arp 102B with photoionization models and infer a density of at least 1011cm−3and a
distance of less than 7 × 1016cm.
The global covering fraction is also difficult to constrain directly from the quasar spec-
– 3 –
trum; we know the gas, at least, partially covers our line of sight, but we have little infor-
mation about other lines of sight. Covering fraction constraints are generally made based
on population statistics. In a seminal paper, Weymann et al. (1991) showed that for most
BALQSOs, the emission line properties are remarkably similar to non-BALQSOs. Thus, the
fact that 10–20% of quasar spectra contain broad absorption lines is interpreted as evidence
that there is a wind that covers 10–20% of sight lines to all similar quasars, and whether or
not we see absorption lines depends on our orientation. Alternatively, some BAL quasars
have notably different line emission than the average quasar; examples are the low-ionization
BALQSOs studied by e.g., Boroson & Meyers (1992). These objects may instead represent
an evolutionary stage of quasars, as the quasar emerges from the cloud of gas and dust in
which it formed (Becker et al. 1997).
While it seems that the column density should be easy to constrain, more recent work
has shown that it can be very difficult to measure. It was originally thought that non-
black absorption troughs indicated a relatively low column density for the absorbing gas
(equivalent hydrogen column densities of 1019−20cm−2, e.g., Hamann 1998). But it has now
been found that the non-black troughs indicate velocity-dependent partial covering, where
the absorption covers part of the emission region, and the uncovered part fills in the trough
partially (e.g., Arav et al. 1999). Thus, the column density appears to be high, but it is very
difficult to constrain directly from the data except in a few very specialized cases (see for
example Gabel et al. 2006; Arav et al. 2005).
How can we make progress on this problem? It is becoming clear that because of the
difficulties described above, the traditional techniques for analysis of troughs (e.g., curve of
growth) and modeling (e.g., photoionization modeling to produce absorption line ratios and
equivalent widths) are limited. An approach that may be profitable is to construct a physical
model for the outflow, and constrain the parameters of the model using the data.
Our first foray into constructing physical models for quasar winds was performed by
Branch et al. (2002). In that paper, the FeLoBAL1FIRST J121442+280329 was modeled
using SYNOW, a parameterized, spherically-symmetric, resonant-scattering, synthetic spec-
trum code more typically used to model supernovae (Fisher 2000). The difference between
this treatment and a more typical one applied to the same data by de Kool et al. (2002) is that
SYNOW assumes that emission and absorption are produced in the same outflowing gas. In
contrast, the approach taken by de Kool et al. (2002) assumes that absorption is imprinted
upon a typical continuum+emission line quasar spectrum; that is, the absorbing gas is sep-
1FeLoBALs are distinguished by the presence of absorption in low-ionization lines such as Al III and
Mg II as well as absorption by excited states of Fe II and Fe III.
– 4 –
arated from the emission-line region. In fact, based on the analysis of the Fe II metastable
absorption lines, they find that the absorber is 1–30 parsecs from the central engine, much
farther than the quasar broad emission-line region. Note that FIRST J121442+280329 is
not the only object that can be modeled using SYNOW; Casebeer et al. (2004) present a
SYNOW model of another FeLoBAL, ISO J005645.1−273816.
The SYNOW model is attractive because it is simple; only one component is needed to
model both the emission and absorption lines. However, this model is limited. The primary
purpose of the SYNOW program is to identify lines in complicated supernova spectra. Thus,
individual ions can be added to a SYNOW run at will in order to see if features from emission
and absorption from those ions is present. It does not solve the physics of the gas, so physical
parameters beyond the existence of a particular species and its velocity extent cannot be
extracted from the results.
We test the ideas of Branch et al. (2002) and Casebeer et al. (2004) by using the
generalized stellar atmosphere code PHOENIX to model the spectra of the two FeLoBALs that
were successfully modeled using SYNOW, and including spectra that extend to rest-frame
optical wavelengths for FIRST J121442+280329. PHOENIX is a much different code than
SYNOW in that it contains all the relevant physics to determine the spectrum of outflowing
gas. It solves the fully relativistic NLTE radiative transfer problem including the effects
of both lines and continua in moving flows. For a discussion of the use of both SYNOW and
PHOENIX in the context of modeling supernovae spectra, see Branch, Baron, & Jeffery (2003).
We find that PHOENIX is able to model the spectra from these objects surprisingly well, and
we are able to derive several important physical parameters from the model.
In §2 we describe the PHOENIX model in detail. In §3 we describe our determination of
the best-fitting model. In §4 we describe the results of our model fitting. In §5 we discuss
the physical implications of the model, how it relates to other BAL spectra and where it fits
in the BAL picture. An appendix includes a flowchart of a PHOENIX calculation.
Photoionization codes have been essential in understanding emission and absorption
features in the spectra of active galaxies. Cloudy (Ferland et al. 1998; Ferland 2003) is widely
used. PHOENIX solves the radiative transfer equations exactly, and tries to have the most
accurate radiative data possible. As a radiative transfer code, PHOENIX correctly produces
line profiles due to relativistic differential expansion. It is this feature that we are making
use of here.
– 5 –
In this paper we model the situation envisioned by Branch et al. (2002): the emission
and absorption occur in a fairly optically thick expanding shell illuminated from the inside by
the continuum; as discussed in §5.4, this situation may be a consequence of quasar evolution,
occurring when the quasar ejects a shroud of dust and gas (e.g., Voit et al. 1993).
PHOENIX is a spectral synthesis code; the direct output is a model spectrum. The only
way to obtain fluxes or equivalent widths of lines in a PHOENIX model is to measure them
directly from the synthetic spectrum in the same way that they are measured from the
observed spectrum. Measuring emission and absorption lines from complex quasar spectra
is well known to be rather uncertain, as a consequence of blending and uncertain placement
of the continuum. So in PHOENIX, this step is bypassed, and the synthetic spectrum is
compared directly with the observed spectrum. Second, the input parameters are somewhat
different. In PHOENIX, density is given as a function of the radius in concordance with the
assumed velocity profile as a function of radius. An analogy to the photoionizing flux is a
little difficult to construct. As noted in the next section, two of the important parameters
are the reference radius R0, the radius at which the continuum optical depth at 5000˚ A is
unity, and the model temperature Tmodel, defined in terms of the total bolometric luminosity
in the observer’s frame, L, and the reference radius. Thus, L or T are somewhat analogous
to the photoionizing flux, because for a fixed reference radius, they give the intensity of
the continuum at the reference radius. Finally, the column density can be evaluated for
particular values of the optical depth.
2.1. The Model Parameters
Our models are spherically symmetric, with homologous expansion (v ∝ r). Homologous
expansion is analogous to the Hubble expansion. The model atmospheres are characterized
by the following parameters (see Baron et al. 2004, for details): (i) the reference radius
R0, the radius at which the continuum optical depth in extinction (τstd) at 5000˚ A is unity;
(ii) the model temperature Tmodel, defined by the luminosity, L and the reference radius, R0,
ve, [ρ(v) ∝ e−v/ve)]; (iv) the expansion velocity, v0, at the reference radius; (v) the pressure,
Pout, at the outer edge of the atmosphere; (vi) the LTE-line threshold ratio, equal to 5×10−6;
(vii) the albedo for line scattering (metal lines only, here set to 0.95); (viii) the statistical
velocity ζ = 50 km s−1, treated as depth-independent isotropic microturbulence, and (ix)
the elemental abundances, assumed to be solar as given by Grevesse & Noels (1993).
0σ))1/4], where σ is Stefan’s constant; (iii) the density structure parameter
We emphasize that for extended model atmospheres one should not assign, a priori,
a physical interpretation to the parameter combination of Tmodel and R0. While Tmodel