Quasars and Cosmology

ASP Conference Series, Vol. 162, 1999

Gary Ferland and Jack Baldwin, eds.

The spectrum of a single photoionized cloud

Gary Ferland
CITA, U of Toronto, and Physics, University of Kentucky

Abstract. The emission-line spectrum of a quasar is most likely emitted

by an ensemble of photoionized clouds moving with a variety of veloci-
ties, with a range of densities and distances from the central object. The
state of the art in this field is to first compute the emission from a single
cloud, and then prescribe a mix of clouds to reproduce line intensities,
profiles, and reverberation lags. Here I review the parameters that de-
fine emission from a single cloud and the outstanding problems in qua-
sar emission-line analysis.

1 Introduction
This is a quick primer for those unfamiliar with the basis of photoionization
modeling. More complete discussions are those of Davidson & Netzer (1979),
Osterbrock (1989), and Netzer (1990). We discuss here only the properties of a
single cloud. Several of the following papers go on to the more realistic situa-
tion of an ensemble of clouds.

2 Photoionization equilibrium of a single cloud

The spectrum emitted by a low-density gas irradiated by an energetic contin-
uum is really just a large-scale fluorescence problem. The ionizing radiation’s
energy is initially converted into a photoelectron’s kinetic energy, which is
then shared with other particles in the cloud, and eventually degraded into the
observed emission line or continuum spectrum.
The essential difficulty is that the gas is far from equilibrium – the concept
of temperature has little meaning. This means that the ionization and level
populations are not given by Saha or Boltzmann statistics, but rather by the
balance of microphysical processes. The only temperature that is ever used is
the kinetic temperature of the electrons. This is generally valid because elastic
electron-electron collisions are very rapid, due to their low mass and high ve-
locity, and insures that a Maxwellian velocity distribution is quickly set up. As
a result of these complications the spectrum is best understood by reference to
large-scale numerical simulations. Ferland et al. (1998) describe such calcula-
tions.

2.1 Ionization: photoionization-recombination balance

The balance between processes that ionize species and those that allow it to
recombine determines the level of ionization of the gas. Ionization processes
include direct photoionization of outer shell electrons, photoionization of inner
shells followed by ejection of Auger electrons, collisional ionization, and
charge transfer ionization. Recombination processes include radiative, dielec-
tronic, and three-body recombination, and charge transfer recombination.

147
148 Ferland

Usually the most important of these processes are valence shell photoioni-
zation and a combination of radiative and dielectronic recombination. In
photoionization problems the flux of ionizing photons striking the illuminated
face of the cloud is a fundamental parameter. This is given by
4π Jν
∞
Φ (H ) ≡∫ dν [photons cm-2 s-1] (1)
ν 1 hν

where Jν is the intensity. The photoionization rate [units s-1] for a species with a
typical photoionization cross section <σ> [units cm2] will be ~ Φ(H) <σ>. In
equilibrium this is equal to the recombination rate, ne α(T), where α(T) is the
recombination coefficient. Both σ and α are the results of extensive atomic
physics calculations and tabulations are readily available (see Verner’s Atomic
Data for Astrophysics web site at http://www.pa.uky.edu/~verner/atom.html).
The ionization balance equation is
natom σ Φ (H ) =ne nionα (T ) (2)
and the resulting level of ionization is
nion Φ (H ) σ σ
= ≈U (3)
natom ne α (T ) cα (T )
where c is the speed of light. The last equation introduces the ionization pa-
rameter U, defined as
Φ(H )
U≡ (4)
nH c
U is sometimes used to parameterize the radiation field instead of Φ(H) be-
cause it is directly proportional to the level of ionization and has the advantage
of introducing homology relations among various models. The photon flux
Φ(H) has the benefit of directly exposing the dependence on the separation be-
tween cloud and central source – this is directly measured by reverberation
studies. In practice either U or Φ(H) can be used as a parameter, and for a
given density the level of ionization goes up with each.

2.2 Temperature: heating-cooling balance

Each photoelectron is ejected with a kinetic energy equal to hν – IP, where hν is
the energy of the incident photon and IP the ionization potential of the atom.
The heating per unit volume due to photoionization will be
∞ 4π Jν
G = natom ∫ σ ν {hν − IP}dν [erg cm-3 s-1] (5)
ν1 hν
where natom is the atom’s density. This kinetic energy is rapidly shared with
other free electrons to establish the electron temperature Te.
Cooling is any process that removes kinetic energy from the free electrons.
Usually the most important cooling process is a collision of an electron with an
ion, resulting in the collisional excitation of a bound electron that then decays,
emitting a photon.
This is clearly an intricate and non-linear problem. The heating/cooling
and resulting temperature cannot be determined until the ionization of the gas
is known. The collisional ionization and recombination rates depend on tem-
perature, so the ionization cannot be determined until the electron temperature
 Photoionization models 149

is known. All of this must be solved as a function of depth into the cloud, tak-
ing into account the attenuation of the continuum by gas opacity and the ef-
fects of radiative transport.

2.3 Energy units, the Rydberg

The spectra shown in various papers in this book have either wavelength
(with units microns, nanometers, or Ångstroms) or energy/frequency (units
Hz, eV, keV) as the x-axis. To make matters worse the natural unit of energy
in photoionization problems is the Rydberg. 1 Rydberg is nearly equal to the
ionization potential of hydrogen, 13.6 eV. The formal definition is
2π 2 me q e4
R∞ ≡ 3
= 109737.315 cm −1 . (1)
ch
R∞ is the ionization potential of an infinite-mass nucleus with a charge of one.
Hydrogen itself has a lower ionization potential due to the reduced mass of the
nucleus, so the Rydberg constant for hydrogen itself is
RH = 2.178728 × 10 −11 erg = 13.59842eV = 91.176340nm = 109677.576cm −1
(2)

3 Photoionization model results

The goal of emission line analysis is to deduce the properties of the clouds that
produce the observed spectrum. This section describes the basic parameters
that determine the conditions in a cloud and the resulting spectrum. These
parameters are the shape of the ionizing continuum, the gas’ density nH, the
chemical composition, the column density NH, and either U or Φ(H).
In the following sections we will describe these parameters and explore
their effects on results. We consider deviations from a single fiducial model, a
cloud with a density of nH = 1010 cm-3, an ionization parameter of U = 10-1.5, so-
lar abundances, and a column density of NH = 1023 cm-2. This density and col-
umn density corresponds to a physical thickness of 1013 cm, or a bit under one
Astronomical Unit.
All of the calculations use a simple power-law continuum, fν ∝ ν α . We
really can’t do any better since the best evidence is that we do not directly ob-
serve the same continuum as the clouds (see the discussion by Zheng in this
volume, Netzer 1985, and also Korista et al. 1997). We will focus on continua
with slopes near the centroid of the observed range, α ∼ –1.5. This power law
continuum extends between 1.36 eV and 50 keV. Outside this range the con-
tinuum is assumed to drop rapidly. Continua are described as soft or hard,
depending on the energy of a typical photon. Harder continua have more en-
ergetic photons and a less negative α.
The approach taken for many years was to try to use the observed spectrum
to obtain the parameters of a typical BLR cloud by comparing this spectrum
with detailed numerical simulations such as those presented below. Today we
know that clouds with a broad range of parameters exist within the emission
line regions, so the notion of a single set of parameters has little meaning.
I will show how some line pairs change with model parameters. These lines
were selected based on the discussion in the meeting. I tried to consistently
show the same set of lines in every plot; if a line ratio is not plotted then it
probably did not change with that particular parameter.
150 Ferland

Finally, all of the calculations presented here were done with version 90.05
of the spectral simulation code Cloudy (Ferland et al. 1998; on the web at
http://www.pa.uky.edu/~gary/cloudy). The programs used to generate
these grids of results are available on this site.

3.1 A cloud’s structure

Figure 1 shows the ionization and thermal structure of a cloud with our fidu-
cial parameters.
The structure has a thin layer of highly ionized gas, He+2, C+3, O+5, extend-
ing to a depth of 1011 cm, followed by region of more moderate ionization, H+,
He+, C+2. A H+ - Ho ionization front occurs at bit under 1012 cm. Unlike galac-
tic nebulae, hydrogen does not become predominately neutral after the ioniza-
tion front, but instead remains partially ionized. This is due to penetrating X-
Rays heating the gas, and the resulting trapped Lyα creating enough popula-
tion in the first excited level of H for photoionization by the Balmer continuum

0
log ionization fractions

2.0

temperature/10 K
++ +
He H
He+ Ho

4
-2 T4 1.5

Ho He+
-4 1.0

Heo

0.5
0 C+3 +
log ionization fractions

-2 C+2 O+5
C+2

Co
-4 C+

C+3

9 10 11 12 13

log depth cm
Figure 1 The ionization structure of a typical blr cloud. The density and
ionization parameter were 10-1.5 and 1010 cm-3, and an α = -1.5 power law ion-
izing continuum was assumed. The calculation stopped at a column density
of 1023 cm-2.
 Photoionization models 151

to become important.
The hydrogen ionization front occurs at the point where nearly all hydro-
gen-ionizing photons have been absorbed. It turns out that the column density
to the hydrogen ionization front is a simple function of U. Since each pho-
toionization is followed by a recombination, the number of recombinations
that occur over the depth D to the ionization front is D n2 α. This must equal
the number of ionizing photons that enter this column, Φ(H), so the depth to
the ionization front is D = Φ(H)/n2 α. The gas column density is N = Dn, so the
column density to the hydrogen ionization front is
N Ifront = Uc / α ≈ 10 23 U [cm-2] . (6)
If the cloud’s total column density is large enough for a hydrogen ionization
front to occur then the cloud is said to be radiation bounded or optically thick to
ionizing photons. A low column density cloud, where the gas remains highly
ionized, is said to be matter bounded or optically thin to ionizing photons.
The temperature is also shown in Figure 1. The gas is hottest in the He+2
region, the temperature falls in He+ region, and then falls still more to ~5000K
in Ho region.

3.2 The ionizing continuum

The shape of the ionizing continuum is both a fundamental parameter for the
model cloud and of great interest because of its connection to the nature of the
central object. We restrict ourselves here to the simple the power-law con-
tinua described in section 3.1. Figure 2 shows the effects of changing the slope
of the power law continuum while keeping all other parameters constant.
The equivalent width of Lyα grows larger as the continuum gets harder.
Lyα is largely a recombination line — calculations show that its intensity re-
mains generally within 0.5 dex of its pure recombination value. In this limit
each hydrogen recombination produces one Lyα photon, so the intensity of
Lyα is proportional to the number of ionizing photons that strike clouds. If the
continuum is simply fν ∝ ν-α then the ratio of the line to continuum is given by
∞ να
hν 3 / 4 ∫ dν
I ( Lyα ) EW ( Lyα ) Ω ν1 hν Ω
= ≈ ≈− (3 / 4)−α α −1 (7)
ν fν (1216) 1216 4π α +1
ν 3/ 4 4π
where ν3/4 is the frequency of Lyα (3/4 of a Rydberg) and the integral is over
all ionizing energies. The results shown in Figure 2 lie close to this simple es-
timate.
The covering factor Ω/4π introduced in eq 7 is the fraction of the photons
emitted by the central object that actually strike clouds. We know that this
must be significantly less than unity since quasars have unobscured lines of
sight to the continuum source. Orientation effects are discussed extensively in
other papers, and it is thought that objects not viewed from this preferred di-
rection are highly obscured.
Quasars have typical Lyα equivalent widths of several hundred Ångstroms.
From Figure 2 we see that, for a typical α of –1.5, the covering factor must be
greater than 1/3 if photoionization is to explain the energetics. Several papers
in this conference allude to either this constraint, or a closely related one that
uses the equivalent width of HeII 1640. Finally, in the nebular literature the
152 Ferland

10 2.0

temperature/10,000K
line ratio
1 1.5
temperature
CIV/Lyα

EW(Lyα)/1000

1.0
-2 -1

SiIII]/CIII] power law spectral index

1

CIII]/CIV
line ratio

Al III/CIII]
0.1

NV/CIV
0.01
OVI/CIV

-2 -1

power law spectral index

Figure 2 The form of the power-law continuum between 1 micron and 50

kev was varied over the indicated range. The plot shows several line ratios,
the equivalent width of Lyα, and the temperature averaged over the O++ zone.
line to continuum ratio is referred to as the Zanstra method of determining
continuum shapes (or equivalently, the stellar temperature).
Figure 2 shows that the O++-weighted mean temperature and CIV/Lyα ratio
both increase with increasingly harder continua. This is because the typical
photoelectron energy increases as the continuum grows harder (equation 5) so
there is more heating per photoionization. CIV 1549 is one of the strongest
coolants in a typical cloud. The ratio CIV/Lyα is essentially the cooling per
recombination, equal to the heating per photoionization, and so to the mean
energy of ionizing photons. This line ratio is closely related to the Stoy method
of determining stellar temperatures in the nebular literature.
We can use line-intensity ratios to deduce the shape of the continuum from
the emission-line spectrum. The basic idea is that the intensity of an H or He
recombination line is proportional to the number of ionizing photons with en-
ergies capable of producing that ion. Then the intensity of a HeII recombina-
tion line relative to Lyα is proportional to the ratio of the number of photons
with energies greater than 4 Ryd to the number greater than 1 Ryd.
 Photoionization models 153

Mathews & Ferland (1987) used this approach to find a continuum that
peaked somewhere near 4 Ryd. This shape does not agree with the continuum
directly observed in z∼1.5 objects (Zheng et al. 1997). Korista, Baldwin, & Fer-
land (1997) pointed out that the observed continuum is not luminous enough
to account for the energy in the high-ionization lines. The implication is that
the emission-line clouds do not see the same continuum we do, a conclusion
previously reached by Netzer (1985). This is not surprising since the geometry
is thought to have a cylindrical symmetry, and there is a growing consensus
that we preferentially view AGN from nearly pole-on directions.

3.3 The infrared continuum

The infrared continuum is critical for dense clouds because of the importance
of free-free or bremsstrahlung heating (Ferland et al. 1992). Free-free heating
is the absorption of a photon by an electron – ion pair, with the electron gain-
ing most of the photon’s energy. The cross section is just the bremsstrahlung
opacity, which is proportional to ν--2 n2, with ν the photon’s frequency and n
the gas density. From this we see that free-free absorption is most important
for dense clouds, and for the lowest frequencies.
10000 106

temperature of face
1000
line ratio

EW(Lyα)

105
T(face)
100

Lyα/Hβ

10
line ratio

CIV/Lyα Al III/CIII]
1

CIII/CIII]
0.1 CIII]/CIV OVI/CIV

IR cutoff, microns NV/CIV

0.01
1 10 100 1000

IR cutoff, microns
Figure 3 Dependence of the spectrum on the infrared cutoff.
154 Ferland

1000
2.5

EW(Lyα)

Temperature/10,000K
line ratio
2.0

temperature

100
1.5

Lyα/Hβ

CIV/Lyα

SiIII]/CIII]
line ratio

0.1 OVI/CIV

CIII]/CIV Al III/CIII]

0.01
0.01 0.1 1 10

metallicity/solar

Figure 4 Line ratios and mean temperature as a function

of the metals scaling factor. The α = -1.5 power law contin-
uum, with a density and ionization parameter of 1010 cm-3
and –1.5 were used.

Figure 3 shows a series of calculations in which the IR break of our fiducial

α = -1.5 power-law continuum was varied. Other cloud parameters were un-
changed. (The continuum used in all other calculations had an IR break at
0.912 micron to minimize the importance of free-free heating.) The figure
shows several line ratios and the temperature of the illuminated face of the
cloud.
For this set of cloud parameters free-free heating becomes dominant when
the UV power law extends to ~30 microns. For denser clouds the effects be-
come more dramatic for much shorter wavelength cutoffs. For instance at a
density of 1012 cm-3 free-free heating becomes important when the cutoff is un-
der 1 micron. This is an additional uncertainly since it is not clear exactly
where the observed IR continuum forms. There are dust and starlight contri-
butions that must originate well beyond the BLR. But there are also cases
where the IR is a component of the central non-thermal continuum (see the re-
 Photoionization models 155

view by Wilkes in this book) and so should be included in the radiation field
striking the clouds.
There has been surprisingly little exploration of the effects of the IR con-
tinua upon BLR clouds, despite the fact that this is more important than the ob-
served hard x-rays. For instance, could differences in the near to far IR contin-
uum, as seen by the BLR, be a contributor to differences in spectra of radio
loud and radio quiet objects?

3.4 Composition
The emission-line spectrum is surprisingly insensitive to the gas composi-
tion. Figure 4 shows a series of calculations in which the abundances of all
elements heavier than helium were varied by the scale factor shown as the x-
axis. Unity represents solar abundances. The CIV 1549/Lyα intensity ratio
changes by only a factor of four while the C/H ratio changes by 3.5 dex. This
reflects energy conservation – CIV is a strong coolant, and the total cooling

1000 20000
EW(Lyα)

temperature
line ratio

temperature
100 15000
Lyα/Hβ

10 10000
CIII]/CIII

1 CIV/Lyα
line ratio

SiIII]/CIII]

CIII]/CIV
0.1
[OIII] 4363/Lyα

AlII/Lyα
0.01
7 8 9 10 11 12 13

log density
Figure 5 Spectrum as a function of density. The calculations
assumed an α=-1.5 power law, solar abundances, and stopped at
the hydrogen ionization front. The lines plotted are Lyα 1216, Hβ
4861, [OIII] 4363, Al III 1857, CIV 1549, CIII] 1909, and SiIII] 1892.
156 Ferland

must balance heating. As the metals scaling factor goes down the cooling effi-
ciency of the gas does too, so the O++-weighted mean temperature (also plot-
ted) goes up. This thermostat effect ensures that the overall spectrum is largely
unchanged despite global changes in the chemical enrichment.
Increasing the abundances of the heavy elements also raises the gas opacity.
At high abundances elements other than H and He absorb ionizing radiation
and the intensity of Lyα relative to the continuum goes down. We have ar-
gued that this is a partial contributor to the Baldwin effect (Korista et al. 1998).
The first studies of quasar spectra found that the abundances are broadly
consistent with solar (see Davidson and Netzer 1979), a conclusion that re-
mains valid today. Although the thermostat effect prevents us from measur-
ing absolute abundances relative to hydrogen, it is possible to measure relative
abundances, such as N/C or N/O. Beginning with Shields (1976), most stud-
ies try to measure the metallicity by measuring the abundance of nitrogen rela-
tive to C and O and then relating this to overall enrichment scenarios. Our re-
cent work suggests that there is a luminosity – metallicity correlation, with the
most luminous objects having a metallicity Z that is 5 – 10 times solar. This is
discussed extensively in Fred Hamann’s chapter in this book, as well as in
Hamann & Ferland (1999).

3.5 Density
The emissivity of a line, the energy released per unit volume and time, will be a
function of the gas density that depends on the detailed atomic physics. The
balance equation for a two-level atom can be written as a balance of two rates
[with unit cm-3 s-1]:
lower to upper = upper to lower
(8)
nl ql ,u ne = nu ( Au ,l β + qu ,l ne )
where β is the escape probability and q is the rate coefficient for collisional pro-
cesses (cm3 s-1). Note that β ~ τ-1, where τ is the line’s optical depth (Elitzur
1992). The emissivity will be
ql ,u ne
ε coll = nu Au ,l β hν = nl Au ,l β hν [ergs cm-3 s-1]. (9)
(Au ,l β + qu ,l ne )
This has a powerful temperature dependence since the rate coefficient ql,u var-
ies as exp(-hν/kT).
The critical density ncrit for a line is the density where the two terms in the
denominator of equation 9 are equal:
ncrit = Au ,l β qu ,l ≈ Au ,l qu ,lτ . (10)
Physically ncrit is the density where the upper level is as likely to be collision-
ally de-excited as to emit a photon. Forbidden lines have ncrit ~ 103 – 106 cm-3,
intercombination lines like CIII] 1909 have ncrit ~ 109 – 1011 cm-3, and permitted
lines ncrit ~ 1014 – 1016 cm-3. Optical depths have the effect of lowering the criti-
cal density since the escape probability multiplies A. Below the critical den-
sity the emissivity varies as the square of the density, and above ncrit, linearly.
Table 1 (based on Baldwin et al. 1996) gives a list of the more prominent ultra-
violet emission lines of quasars along with an indication of their formation
mechanism and critical density.
 Photoionization models 157

Figure 5 shows the results of changing the density of our standard cloud
while keeping the ionization parameter constant. The temperature grows hot-
ter as n increases because denser gas does not cool very efficiently. This is be-
cause many of the strong coolants are suppressed above their ncrit. At the
highest densities the cloud has significant free-free opacity at IR wavelengths,
and free-free absorption becomes a major heating agent, even with our IR
break at 0.912 microns. Lyα becomes weaker relative to the continuum at high
densities, as the line becomes thermalized, i.e. approaches the black body limit
for its emission. Very dense clouds are sources of continuum, not line emis-
sion. The lower panel of Figure 5 shows several line ratios that have been
used as density indicators. Basically, as all lines become thermalized, their rela-
tive intensities approach 1:1 ratios, as each emits near the black body limit for
its temperature and wavelength. Note the great strength of Al III] and SiIII]
relative to C III] at high densities. Baldwin et al. (1996) argue that these lines
are density indicators. The Lyα/Hβ ratio is also small at high densities.

Table 1 Prominent lines in quasars

Ion λ(A) Components Formed by log ncrit
C III 977 977.03 rec, coll, pmp 16.2
N III 991 990.98 rec, coll, pmp 15.9
Lyβ 1026 1025.72 rec 15.4
O VI 1034 1031.95, 1037.63 coll 15.7
Lyα 1216 1215.67 rec, coll 16.9
NV 1240 1238.81, 1242.8 coll 15.5
Si II 1263 1260.42, 1264.73 ?? 16.4
OI 1303 1302.17, 1304.87, 1306.04 pmp, coll 17.0
Si II 1307 1304.37, 1309.28 ?? 15.9
C II 1335 1334.53, 1335.66, 1335.71 coll 15.7
Si IV 1397 1393.76, 1402.77 coll 15.6
O IV] 1402 1397.21, 1399.78, 1404.79, 1407.39 coll 11.0
N IV] 1486 1486.5 coll 10.2
C IV 1549 1548.20, 1550.77 coll 15.3
He II 1640 1640.72 rec 16.1
O III] 1665 1660.80, 1666.14 coll 10.5
N III] 1750 1748.65, 1752.16, 1754.00 coll 10.3
Fe II UV191 1786.7 ?? 16.4
Si II 1814 1808.00, 1816.92, 1817.45 coll 13.4
Al III 1857 1854.72, 1862.78 coll 15.4
Si III] 1892 1892.03 coll 11.0
C III] 1909 1908.73 coll 9.5
Fe III UV 34 1895.46, 1914.06, 1926.30 ?? 16:
Fe III UV 48 2061.55, 2068.24, 2078.99 ?? 16:
N II] 2142 2139.7, 2143.5 coll 10.0
Mg II 2798 2795.53, 2802.70 coll 15.0
158 Ferland

7
3 Compton
thick

log line ratio or EW

EW(Lyα)

log temperature
6
2
Lyα/Hβ

5
1
temperature
radiatively
unstable 4
0
CIII]/CIV
1
CIII]/CIII OVI/CIV
log line ratio

0
NV/CIV
CIV/Lyα
-1
SiIII]/CIII]
Al III/CIII]

-2
-4 -3 -2 -1 0 1 2 3

log ionization parameter

Figure 6 Spectrum as a function of the ionization parameter. Values of the
ionization parameter where the cloud is optically thick to electron scattering
are indicated as “Compton thick”, and where line radiation pressure exceeds
gas pressure by “radiatively unstable”.

3.6 Flux of ionizing photons or ionization parameter

Figure 6 shows the effects of changing the ionization parameter. In these cal-
culations the column density was not held fixed, but the calculation was
stopped at the point where the gas temperature fell below 4000 K. The tem-
perature at the illuminated face of the cloud is shown in the upper panel of
Figure 6. Both the O++-weighted mean T (plotted) and ionization goes up with
increasing U.
The ionized column density increases with U (equation 6), and eventually
the column density is large enough for the cloud to become optically thick to
electron scattering. The figure shows the point where τe > 1 and also where the
internally generated line radiation pressure exceeds the gas pressure. These
clouds would be unstable to disruption by radiation pressure if they are exter-
nally supported, such as would happen with a hot intercloud medium (Elitzur
and Ferland 1986).
 Photoionization models 159

10 50

Lyα/Hβ NIII/NIII]
45

line ratio

Lyα/Hβ
CIII/CIII]
40
1

35

EW(Lyα)/1000

10

NIII/NIII]

CIII/CIII]
1
line ratio

Al III/CIII] Sil III]/CIII]

0.1

OVI/CIV
CIII]/CIV
NV/CIV

0.01
10 100 1000 10000

turbulent velocity
Figure 7 The effects of microturbulent velocity.
For low values of the ionization parameter the level of ionization of the gas,
as evidenced by ratios of lines from different ions of the same element, goes up
with U. This prompted the early statements that the spectrum depended
strongly on U. Figure 6 shows that the CIII]/CIV ratio is indeed strongly sensi-
tive to U for U < 10-1.5. Actually, for high values of U the column density of the
ionized part of the cloud is so large that individual elements form their own
Strömgren shells. In this limit the line ratios are insensitive to U and become
more sensitive to the continuum shape.
There has been almost no exploration of these very large U solutions. The
fact that large U clouds can be Compton thick brings in electron scattering as a
additional line-broadening mechanism. This could help explain a long-
standing puzzle – the profiles of broad emission lines remain smooth out into
the line wings. If lines from each cloud are only thermally broadened then this
smoothness requires an impossibly large number of clouds (Arav et al. 1998).
A component of electron broadening would make each cloud’s intrinsic line
spectrum exceptionally fuzzy and so help this dilemma.
160 Ferland

3.7 Turbulent velocity

From the beginning, analyses of clouds have assumed that each cloud has
only thermal motions, and that the observed very broad line profiles are due to
macroscopic motions of the clouds. There has been very little exploration of
this assumption.
Figure 7 shows the effects of adding a component of turbulence to our
standard cloud. As the line width increases there are two main effects. The
first is to decrease line optical depths and so allow resonance lines to escape
more freely. They then cool the gas more efficiently. The second is to increase
the efficiency of line pumping by the incident continuum. The gas tempera-
ture (not plotted) tends to go down, as trapped resonance lines become more
efficient coolants at large values of the turbulence. The greatest single effect is
to increase the intensities of far UV lines (those shortward of Lyα) by adding a
pumped component to the (small) component due to collisional excitation.

4 Putting it all together - the curse of the free parameter

The previous discussion should have been more than a little depressing –
we have less than a dozen lines whose intensities can be measured in a typical
spectrum, and over a dozen pages of free parameters! The miracle is that it is
still hard to fit the spectrum.
An approach to understanding the spectrum is suggested by the series of
plots shown so far. We have a finite number of free parameters, with the den-
sity and flux/ionization parameter having the most powerful effects on the
spectrum, and the composition and shape of the ionizing continuum being the
most astrophysically interesting. The next step is to isolate which pairs of lines
are most sensitive to which parameters – are there line pairs that tell us the
density? Baldwin et al. (1996) argue that Al III/C III] is one pair. Others in
this volume argue for Si III]/C III], which seems well correlated with Eigen-
vector 1. Lines of nitrogen relative to C and O lines can tell us the abundances
if we think we understand the chemical enrichment history of the environment
(Hamann, this volume). The HeII spectrum tells us about the incident contin-
uum at unobservable wavelengths.
Quantitative spectroscopy of emission line sources was identified as a major
area of research very early in the history of computational astrophysics
(Hjellming 1966, Rubin 1968, Williams 1967, Davidson 1972; MacAlpine 1971).
Any one of the figures shown in this paper would have required a large in-
vestment of computer time at a major research university in the early 1970’s,
and it probably would not have been possible for an investigator to do them
all. This prevented a complete reconnaissance of parameter space. There are
still regimes of parameter space that have not been well explored, as described
in previous sections above.
The biggest question for the original workers was why the spectra of qua-
sars are so similar – they do not display the range of line intensities shown by
planetary nebulae, for instance. This was taken as an indication that the cloud
parameters, especially U and n, were always the same. Much of the inital
work went into efforts to identify the agents responsible for fine-tuning the
parameters. The hot – cold model proposed by Krolik, McKee, & Tarter (1981)
was one way to select parameters, but the postulated hot phase has so much
opacity that little radiation would escape (Mathews & Ferland 1987). The even
 Photoionization models 161

400

300

EW (CIV)
200

100

19
log
20
Φ(

21 8
9
H)

10
22
11
12
y
23 13
g d ensit
lo
Figure 8 The predicted equivalent width of CIV 1549.
softer continuum measured by Zheng et al. (1997) makes this even worse. It is
very hard to support the existence of a pervasive hot phase.
With the power of modern workstations it is now possible to explore far
more of parameter space. Figure 8 gives an example, based on Baldwin et al.
(1995). It shows the equivalent width of CIV 1549 as a function of the density
and flux of ionizing photons. This illustrates the fact that CIV is produced ef-
fectively by only a narrow range of parameters, as discussed extensively in the
chapter by Kirk Korista in this book. This lead to the LOC approach outlined
by Korista, which uses selection effects alone, with no preferred parameters, to
fit the observed spectrum. If it turns out that this picture is correct then pa-
rameters of individual clouds can be ignored and we can use the clouds to
probe more fundamental properties of quasars such as metallicity and the ef-
fects of the SED.

5 Outstanding problems
There have been great advances in numerical simulations of quasar clouds,
largely the result of the power of modern computers. But there are still major
questions left unanswered. These are:
♦ The Lyα/Hβ ratio has not really been solved. Baldwin (1977) discovered that
this ratio is under 10, but all of the plots shown in this paper predict be-
tween 30 and 100 (see also Netzer et al. 1995). Several solutions are pos-
sible – the first is that the spectrum has been reddened. Another is that
the radiative transfer in the current generation of plasma simulation
codes is not good enough. Another is suggested by the Kwan and Kro-
162 Ferland

Krolik (1980) work, which did reproduce the observed ratio but with an
ionizing continuum that is far harder than is found in AGN – perhaps
the actual continuum striking clouds is indeed very hard. Finally, low
values of Lyα/Hβ can be produced by LOC integrations that extend to
very high densities (Baldwin 1997).
♦ The BLR is highly stratified. This is shown by reverberation studies (see
the review by Horne in this book). Clearly the type of approach taken
here is too simplistic – we must be thinking about the global environ-
ment and a range of cloud properties. Rees, Netzer, and Ferland (1989)
assumed that the clouds were controlled by an external pressure, which
they approximated by a simple power law. Kaspi and Netzer discuss
this type of model in a chapter in this book. Another approach is the
LOC described by Korista in this book. Here clouds have a very broad
range of properties and selection effects introduced by the atomic phys-
ics controls the observed spectrum.
♦ High and low ionization lines do not have the same redshift (Gaskell 1982;
Wilkes 1986; Espey et al. 1989). This is telling us something about the
BLR velocity field in a stratified environment. There is not any one
model for the origin of the shifts that is accepted, nor any that really
works.
♦ Why are emission line profiles so smooth? If individual clouds have ther-
mal line widths and the observed line widths are due to bulk motion of
a large numbers of clouds, the profiles should break up into individual
components far out in the wings were Gaussian statistics say that few
clouds should contribute to the profile. This does not happen, leading
to the conclusion that an impossibly large number of clouds are in-
volved (Capriotti, Foltz, and Byard 1981; Arav et al. 1998). This could
be explained with either electron scattering by relatively cool gas (T <
106 K) or microturbulence.
♦ Why equivalent width? Current models of the Baldwin effect assume that
it is driven by changes in the continuum shape, perhaps combined with
metallicity. If this were all that is going on we would expect line ratios to
be better correlated with luminosity, and so have less scatter, than a line
relative to the continuum. The line equivalent width brings in an addi-
tional parameter, the covering factor (eqn 7), which will introduce addi-
tional noise unless Ω/4π is always the same. Additionally the contin-
uum will be beamed differently than that of the lines. We have looked
for emission-line ratio correlations with luminosity but found none bet-
ter than the equivalent width correlations. This is unexpected from the
nebular physics.

I thank Jack Baldwin and Kirk Korista for their help and comments. NSF
and NASA support research in Nebular Astrophysics at the University of Ken-
tucky.

6 References
Arav, N., Barlow, T.A., Laor, A., Sargent, W., & Blandford, R., 1998, MNRAS 297, 998
Baldwin, J.A., 1977, MNRAS 178, 67
Baldwin, J.A. 1997, in Emission Lines in Active Galaxies: New Methodsand Tech-
niques, ed. B.M. Peterson, F.-Z. Cheng & A.S. Wilson (ASPConference Ser., 113),
80
 Photoionization models 163

Baldwin, J. A., Ferland, G.J., Korista K. T., and Verner, D., 1995ApJ 455, L119
Baldwin J.A., Ferland, G.J., Korista, K.T., Carswell, R., Hamann, F., Phillips, M.,
Verner, D., Wilkes, B., & Williams, R.E., 1996, ApJ 461, 683
Capriotti, E., Foltz, C.B., and Byard, P., 1981, ApJ 245, 396
Davidson, K., 1972, ApJ 171, 213
Davidson, K., & Netzer, H., 1979, Rep. Prog. in Physics 51, 715
Elitzur, M, 1992, Astronomical Masers, Kluwer, Dordrecht
Elitzur, M., & Ferland, G.J., 1986, ApJ 305, 35
Espey, B., et al. 1989, ApJ 342, 666 erratum 354, 763
Ferland, G.J., Peterson, B.M., Horne, K., Welsh, W.F., & Nahar, S.N., 1992, ApJ 387, 95
Ferland, G. J. Korista, K.T. Verner, D.A. Ferguson, J.W. Kingdon, J.B. Verner, & E.M.
1998, Publ. A.S.P., 110, 761
Gaskell, C.M., 1982, ApJ 263, 79
Hamann, F., & Ferland, G., 1999, ARAA in press
Hjellming, R.M., 1966, ApJ 143, 420
Korista, K.T., Baldwin, J., and Ferland, G.J., 1998, ApJ 507, 24
Korista, K.T., Ferland, G.J., and Baldwin, J., 1997, ApJ 487, 555
Krolik, J., McKee, C.M., & Tarter, C.B., 1981, ApJ 249, 422
Kwan, J., & Krolik, J., 1981, ApJ 250, 478
Mathews, W.G., & Ferland, G.J., 1987, ApJ 323, 456
MacAlpine, G. M., 1971, ApJ 175, 11
Netzer 1985 ApJ 289, 451
Netzer, H., 1990, in Active Galactic Nuclei, Saas-Fee Advanced Course 20, Courvorsier,
T.J.-L., & Mayor, M., (Springer-Verlag; Berlin)
Netzer, H., et al. 1995, ApJ 448, 27
Osterbrock, D. E., 1989, Astrophysics of Gaseous Nebulae & Active Galactic Nuclei, Univer-
sity Science Press
Rees, M.J., Netzer, H., & Ferland, G.J., 1989, ApJ 347, 640
Rubin, R.H., 1968, ApJ 153, 671
Shields, G.A., 1976, ApJ 204, 330
Wilkes, B., 1986, MNRAS 218, 331
Williams, R.E., 1967, ApJ 147, 556
Zheng, W., et al. 1997 ApJ 475, 469