ArticlePDF Available

Abstract and Figures

We briefly review, at a level appropriate for graduate students and non-specialists, the field of quasar absorption lines (QALs). Emphasis is on the intervening absorbers. We present the anatomy of a quasar spectrum due to various classes of intervening absorption systems, and a brief historical review of each absorber class (Lyman-alpha forest and Lyman limit systems, and metal-line and damped Lyman-alpha absorbers). We also provide several heuristic examples on how the physical properties of both the intergalactic medium and the gaseous environments associated with earlier epoch galaxies can be inferred from QALs. The evolution of these environments from z=4 are discussed. Comment: 15 pages, 9 figures; Written for the Encyclopedia of Astronomy and Astrophysics (to be published in 2000 by Mac Millan and the Institute of Physics Publishing)
Content may be subject to copyright.
arXiv:astro-ph/0006002v1 31 May 2000
Quasistellar Objects: Intervening
Absorption Lines
1
Jane C. Charlton and Christopher W. Churchill
The Pennsylvania State University, University Park,
PA 16802
Abstract
We briefly review, at a level appropriate for grad-
uate students and non-specialis ts, the field of quasar
absorption lines (QALs). Emphasis is on the interven-
ing absorbers. We present the anatomy of a quasar
sp e c trum due to various classes of intervening absorp-
tion systems, and a brief historical review of each ab-
sorber class (Lyman-alpha forest and Lyman limit
systems, and metal-line and damped Lyman-alpha
absorbers). We also provide several heuristic exam-
ples on how the physical properties of both the inter-
galactic medium and the ga seous environments asso-
ciated with earlier epoch galaxies can be inferre d from
QALs. The evolution of these environments from z=4
are discussed.
1. Introduction
Every parcel of gas along the line of sight to
a distant quasar will selectively absorb certain
wavelengths of continuum light of the quasar due
to the presence of the various chemical elements
in the gas. Through the analysis of these quasar
absorption lines we can study the spatial distri-
butions, motions, chemical enrichment, and ion-
ization histories of gaseous structures from red-
shift five until the present. This includes the gas
in galaxies of all morphological types as well as
the diffuse gas in the intergalactic med ium.
1.1. Basics of Quasar Spectra
Figure 1 illustrates many of the common fea-
tures of a quasar spectrum. The relatively flat
1
Written for the Encyclopedia of Astronomy and Astro-
physics (to be published in 2000 by MacMillan and th e
Institute of Physics Publishing)
quasar continuum and broad emission features
are produced by the quasar itself (near the black
hole and its accretion disk). In some cases, gas
near the quasar central engine also produces “in-
trinsic” absorption lines, m ost notably Lyα, and
relatively high ionization metal transitions such
as Civ, Nv, and Ovi. These intrinsic absorp-
tion lines can be b road [thousands or even tens
of thousands of km s
1
in which case the quasar
is called a broad absorption line (BAL) QSO], or
narrow (tens to hundreds of km s
1
). However,
the vast majority of absorption lines in a typi-
cal quasar spectrum are “intervening”, produced
by gas unrelated to the quasar that is located
along the line of sight between the quasar and
the Earth.
A structure along the line of sight to the
quasar can be described by its neutral Hydrogen
column density, N(Hi), the number of atoms per
cm
2
. N(Hi) is given by the product of the den-
sity of the material and th e pathlength along the
line of sight through the gas. Each structure will
produ ce an abs orption line in the quasar spec-
trum at a wavelength of λ
obs
= λ
rest
(1 + z
abs
),
where z
abs
is the redshift of the absorbing gas
and λ
rest
= 1215.67
˚
A is the rest wavelength
of th e Lyα transition. Since z
abs
< z
QSO
,
the redshift of the quasar, th ese Lyα absorp-
tion lines form a “forest” at wavelengths blue-
ward of the Lyα emission. The region redward
of the Lyα emission will be populated only by
absorption through other chemical transitions
with longer λ
rest
. Historically, absorption sys-
tems with N (Hi) < 10
17.2
cm
2
have been called
Lyα forest lines, those w ith 10
17.2
< N (Hi) <
10
20.3
cm
2
are Lyman limit systems, and those
with N(Hi) > 10
20.3
cm
2
are damped Lyα sys-
tems. The number of systems per un it redshift
increases dramatically with decreasing column
density, as illustrated in the schematic diagram
in Figure 2. Lyman limit systems are defined
by a sharp break in the spectrum due to absorp-
tion of photons capable of ionizing Hi, i.e. those
with energies greater than 13.6 eV. The optical
1
1600 1800 2000 2200 2400 2600 2800 3000 3200
0
1
2
3
4
5
Fig. 1.— Typical spectru m of a quasar, showing the quasar continuum and emission lines, and the absorption lines
produced by galaxies and intergalactic material that lie between the q uasar and the observer. This spectrum of t he z = 1.34
quasar PKS0454 + 039 was obtained with the Faint Object Spectrograph on the Hubble Space Telescope. The emission
lines at 2400
˚
A and 2850
˚
A are Ly β and Ly α. The Ly α forest, absorption produced by various intergalactic clouds,
is apparent at wavelengths blueward of the Lyα emission line. The two strongest absorbers, due to galaxies, are a damped
Lyα absorber at z = 0.86 and a Lyman limit system at z = 1.15. The former produces a Lyman limit break at 1700
˚
A
and the latter a partial Lyman limit break at 1950
˚
A since the neutral Hydrogen column density is not large enough for it
to absorb all ionizing photons. Many absorption lines are produced by the DLA at z = 0.86 (Civ λλ1548, 1550, for example,
is redshifted onto the red wing of the qu asar’s Ly α emission line).
depth, τ, of the break is given by the product
N(Hi)σ, where th e cross section for ionization of
Hydrogen, σ = 6.3 × 10
18
(E
γ
/13.6 eV)
3
cm
2
,
(and the fl ux is reduced by the factor e
τ
). The
energy depend en ce of σ leads to a recovery of
the Lyman limit break at higher energies (shorter
wavelengths), unless N(Hi) 10
17.2
cm
2
(see
Figure 1).
The curve of growth describes the relationship
between the equivalent w idth of an absorption
line, W , (the integral of the normalized p rofile)
and its column density, N. Figure 3 shows that
for small N (Hi) the number of ab s orbed pho-
tons, and therefore the flux removed, increases
in direct proportion to the number of atoms.
This is called the linear part of the curve of
growth. As N is increased the line saturates so
that photons are only absorbed in the wings of
the lines; in th is r egime the equivalent width is
sensitive to the amount of line broadening (char-
acterized by the Doppler parameter b), but does
not depend very strongly on N(Hi). This is
the flat part of the curve of growth. Finally, at
N(Hi) > 10
20.3
cm
2
, there are enough atoms
that th e damping wings of the line become pop-
ulated and the equivalent width increases as the
square root of N(Hi), and is no longer sensitive
to b.
In addition to the Lyα (1s 2p) and higher
order (1s np ) Lyman series lines, quasar spec-
tra also show absorption due to different ioniza-
tion states of the various species of metals. Fig-
ure 1 illustrates th at the damped Lyα system at
z = 0.86 that is responsible for the Lyα absorp-
tion line at λ
obs
= 2260
˚
A and a Lyman limit
break at λ
obs
= 1700
˚
A also produces absorp-
2
12 14 16 18 20 22
-24
-20
-16
-12
Fig. 2.— The column density distribution of Lyα
clouds, f(N(H i), roughly follows a power law over ten or-
ders of magnitude; there are many more weak lines than
strong lines. The column density regions for the three cat-
egories of systems are shown: Ly α forest, Lyman limit,
and damped Ly α. The term “Lyα forest” has at times
been used to refer to metal–free Hydrogen clouds, perhaps
those with N(Hi) < 10
16
cm
2
, but now metals have been
found associated with weaker systems down to the detec-
tion limit.
tion at λ
obs
= 2870
˚
A due to the presence of
Civ in the absorbing gas at that same redshift.
Like many of the strongest metal lines seen in
quasar spectra, Civ is a resonant doublet tran-
sition due to transitions from
2
S
1/2
energy lev-
els to the
2
P
1/2
and to the
2
P
3/2
energy levels.
(The left superscript “2” rep resents the number
of orientations of the electron spin, the letter S
or P represents th e total orbital angular momen-
tum, L, and the right subscript represents the to-
tal angular momentum, J.) Doub let transitions
are easy to identify. The dichotomy between rest
wavelength and redshift is resolved because the
observed wavelength separation of the doublet
members increases as 1 + z.
Table 1 lists some of the metal lines that are
commonly detected for intervening absorption
systems. Many of these are only strong enough to
be observable for quasar lines of sight that pass
through the higher N(Hi) regions of galaxies.
Table 1: Common Transitions
Transition λ
rest
[
˚
A]
LL . . . . . . . . . . . . . 912
Lyγ . . . . . . . . . . . . 972.537
Lyβ . . . . . . . . . . . . 1025.722
Lyα . . . . . . . . . . . . 1215.670
Siiv 1393 . . . . . . . 1393.755
Siiv 1402 . . . . . . . 1402.770
Civ 1548 . . . . . . . 1548.195
Civ 1550 . . . . . . . 1550.770
Feii 2382 . . . . . . . 2382.765
Feii 2600 . . . . . . . 2600.173
Mgii 2796 . . . . . . 2796.352
Mgii 2803 . . . . . . 2803.531
2. History, Surveys, and Revolutionary
Progress in the 1990’s
The history of quasar absorption lines began
within a couple of years of the identification of
the first quasar in 1963. In 1965, Gunn and
Peterson considered the detection of flux blue-
ward of the Lyα emission line in the quasar
3C 9, observed by S chmidt, and derived a limit
on the amount of neutral Hydrogen that could
be p resent in intergalactic space. In that same
year, Bahcall and Salpeter predicted that inter-
vening material should produce observable dis-
crete absorption features in quasar spectra. Such
features were detected in 1967 in the quasar
PKS 0237 23 by Greenstein and Schmidt, and
in 1968 in PHL 938 by Burbidge, Lynds, and
Stockton. By 1969 many intervening systems
had been discovered, and Bahcall and Spitzer
proposed that most with metals were produced
by the halos of normal galaxies. As more data ac-
cumulated, the sheer numb er of Lyα forest lines
strongly supported the idea that galactic and in-
tergalactic gas, and not only material intrinsic to
the quasar, is the source of most quasar absorp -
tion lines.
In the 1980’s many more quasar spectra were
3
-200-150-100 -50 0 50 100 150 200
0
.5
1
-100 -50 0 50 100
.4
.6
.8
1
1.2
-300 -200 -100 0 100 200 300
0
.5
1
12 13 14 15 16 17 18 19 20 21
-2
-1
0
1
Fig. 3.— Illustration of the d ifferent regimes of the curve of growth. The middle panel shows th e curve of growth for
the Lyα transition, relating the equivalent width, W , of the absorption profile to the column density, N(Hi). The different
curves represent four different values of the Doppler parameter: b = 13, 23, 53, and 93 km s
1
. The upper panel shows
absorption profiles with Doppler parameter b = 23 km s
1
for the series of neutral hydrogen column densities N(Hi) = 10
12
10
20
cm
2
. The thick (thin) curves correspond to the filled (open) points on the b = 23 km s
1
curve of growth (middle
panel), starting at N(Hi) = 10
12
cm
2
. For N(H i) < 10
13
cm
2
, known as the linear part of the curve of growth, the
equivalent width does not depend on b. The lower left panel shows that, at xed N(Hi), the depth of the profile is smaller
for large b, such that the equivalent width remains constant. On the flat part of t he curve of growth, profiles are saturated
and the equ ivalent width increases with b for constant N(H i). For N(Hi) > 10
20
cm
2
, th e p rofile develops damping wings,
which dominate the equivalent width.
4
obtained and many large statistical surveys of
the different classes of absorption line systems
were published. The emphasis was to character-
ize the number of lines per unit redshift, dN/dz,
stronger than some specified equivalent width
limit. With 4m–class telescopes [equipped with
charge coupled device (CCD) detectors] it was
possible to conduct surveys with a spectral res-
olution of R 1000. The spectral resolution is
defined as R = λ/λ = c/v, so that R = 1000
corresponds to 300 km s
1
or 5
˚
A at λ = 5000
˚
A.
Separate surveys were conducted for Lyα lines,
Mgii doublets, Civ d oublets, and also for Ly-
man limit breaks, all as a function of red shift.
The Lyα line is observable in the optical part of
the spectrum for z > 2.2, Mgii for 0.4 < z < 2.2,
Civ for 1.7 < z < 5.0, and the Lyman limit
break f or z > 3. However, a break is also eas-
ily identified in lower resolution space–based UV
spectra, which extended Lyman limit sur veys to
lower redshift.
In order to consider the cross section of the
sky covered by the different populations, it can
be assumed that absorption will be observed for
all lines of sight within some radius of every lu-
minous galaxy (> 0.05L
K
). (L
K
represents the
Schechter luminosity, i.e. the transition between
the expon ential and the power law forms of the
luminosity function, and corresponds to a K
band absolute magnitude of M
K
= 25). To
explain the observed dN/dz at z 1.5, this
radius would be 70 kpc for strong Civ (detec-
tion sensitivity 0.4
˚
A), and 40 kpc for strong
Mgii (detection sensitivity 0.3
˚
A) and also for
Lyman limit sys tems, implying that the latter
two populations are in fact produced in the same
gas. The higher N(Hi) damped Lyα absorbers
would be produced within 15 kpc of the center
of each galaxy, while the Lyα forest lines would
require a considerably larger region, hundreds of
kpcs around each galaxy to produce a cross sec-
tion consistent with the observed number of weak
lines.
Up until the 1990’s, the focus of quasar ab-
sorption line work was to separately consider the
properties of the individual classes of absorbers
(eg. Lyα forest or Mgii absorbers). In the 1990’s,
however, three different observational advances
led to recognition of the direct connections be-
tween the different classes of quasar absorption
lines, and of direct asso ciations with the popula-
tion of galaxies:
1. Deep images of quasar fields could be ob-
tained, and redshifts of the galaxies in the field
could be determined from low resolution spec-
tra. Steidel found that w henever Mgii absorp-
tion with W
r
(Mgii) > 0.3
˚
A is observed, a lu-
minous galaxy (L
K
> 0.06L
K
) is f ound within
an impact parameter of 38h
1
(L/L
K
)
0.15
kpc
with a r ed shift coincident with that determined
from the absorption lines. Also, it is rare to nd
a galaxy within this impact parameter that does
not produce Mgii absorption. There appears to
be a one–to–one correspondence between strong
Mgii absorption and lum inous galaxies. The
Mgii absorbing galaxies span a range of morpho-
logical types.
2. The High Resolution Spectrograph on the
Keck I 10-meter telescope made it possible to ob-
tain quasar spectra at a resolution of R = 45, 000,
which corresponds to 6 km s
1
. The previ-
ous surveys with resolution of order hundreds of
km s
1
identified absorption due to entire galax-
ies and their environments. With 6 km s
1
res-
olution it became possible to resolve structure
within a galaxy: the clouds in its halo, the in-
terstellar medium of its disk, and the satellites
and infalling gas clouds in its environment. Fig-
ure 4 is a dramatic illustration of this contrast for
the Mgii absorber at z = 0.93 toward the quasar
PG 1206 + 459.
3. The Faint Object Spectrograph (FOS) on
the Hubble Space Telescope provided resolution
R 1000 in the UV, from 1400–3300
˚
A. Obser-
vations of Lyα forest clouds could be extended
from z = 2.2 down to the present epoch. Fur-
thermore, absorption from a given galaxy could
be observed in numerous transitions; if Mgii was
5
5390 5400 5410 5420
0
.5
1
1.5
2960 2980 3000 3020
0
.5
1
1.5
5000 5500 6000 6500 7000
0
2000
4000
6000
Fig. 4.— Dramatic demonstration of gains due to h igh resolution spectroscopy of the Mgii doublet. The top panel is
a R = 3000 spectrum of PG1206 + 459. The d oublet that is apparent at an observed wavelength of 5400
˚
A is due to
Mgii absorption from a system at z = 0.927. The middle panel shows the remarkable kinematic struct ure that is revealed
at the resolution (R = 45, 000) of the Keck/HIRES spectrograph of the same quasar. The 2796
˚
A transition is resolved
into multiple components (5583–5592
˚
A), which also appear in the 2803
˚
A transition (5396–5406
˚
A). This system can be
separated in two “clusters of clouds, labeled “A” and “B”. Another weaker Mgii doublet is observed at 5409 and 5423
˚
A,
from a system at z = 0.934
˚
A, labeled with a “C”. The solid line through these complex Mgii profiles is the result of multiple
Voigt profile fitting, with a cloud centered on each of the ticks drawn above the spectrum. The lower panel shows the
Civ doublets associated with the same three systems, observed with the Faint Object Spectrograph on HST, but at much
lower resolution (R = 1300). The C iv is in three different concentrations around the three systems “A”, “B”, and “C”. The
Civλ1550 transition from system A is blend ed with the Civλ1548 transition from system B. The Civ equivalent width is too
large for this absorption to be produced by the same phase of gas that produces the Mgii cloud absorption. The maximum
absorption that can arise in the Mgii phase is given by the dotted line; a plausible model with a kinematically broader Civ
phase yields the solid curve.
6
observed in the optical, the Lyman series and
Civ could be stu died in the UV (see Figure 4).
With information on transitions with a range of
ionization states, consideration of the degree of
ionization (related to the gas density and the in-
tensity and shape of the ionizing radiation field)
and the multiple phase structure of galactic gas
became possible.
No longer is analysis of absorption lines in
quasar spectra an esoteric subject. It h as devel-
oped into a powerful tool to be used in the study
of galaxy evolution (eg. similar to imaging the
stellar components of the galaxies). At least in
principle, quasar spectra can be used for an unbi-
ased study of the gaseous environments of galax-
ies from the present back to the highest redshifts
at which quasars are observed. Gas s tr uctures
smaller than 1 M
can be detected if they are
intercepted by the quasar line of sight, irrespec-
tive of whether they emit light. Through the tool
of quasar absorption lines, proto–galactic struc-
tures and low surface brightness galaxies can be
studied as well as high luminosity galaxies.
3. Developing Physical Intuition
With high r esolution spectra of quasars, it is
possible to consider the p hysical conditions of
the gaseous structures that produce absorption.
However, it is challenging to separate the vari-
ous effects that “shape” the spectral features in
the different chemical transitions. The absorp-
tion profiles observed for the different chemical
transitions are determined by a combination of
the spatial distribution of material along th e line
of sight, its bulk kinematics, temperature, metal-
licity, and abundance pattern. The ionization
structure is influenced by gas densities and by
the UV radiation field, wh ich is a combination
of the extragalactic background radiation due to
the accumulated effect of quasars and stellar pho-
tons escaped from galaxies (and corrected for ab-
sorption by the intergalactic medium).
The shape of an absorption line can be mod-
eled with a Voigt profile, which is a combination
of the natural, quantum mechanical Lorentzian
broadening and the Gaussian broadening caused
by the thermal and turbulent motions in the gas.
Several Voigt profiles can be blended together to
form an overall complex absorption feature (see
Figure 4). The “width” of a single Voigt profile
is characterized by the Dopp ler p arameter, b (ex-
pressed in velocity units and related to the Gaus-
sian σ by b = 2
1/2
σ). Physically, the Doppler
parameter is the sum of thermal and turbulent
components, b
2
tot
= 2k T /m + b
2
turb
, where T is
the temperature of the gas, and m is the mass of
an atom.
3.1. Kinematic Models
Two of the simplest types of organized kine-
matics in galaxies are illustrated in Figure 5:
clouds distributed in a rotating disk, and radial
infall of clouds in a spherical distribution. Here,
Mgii absorbers are used as an example, but the
same kinematic arguments would apply to other
transitions. For radial infall, clouds can be dis-
tributed over the range of velo cities, with a ten-
dency for a “double peak” from material that is
redshifted and blueshifted but with a consider-
able amount of variation if there are typically
several discrete clouds along the line of sight.
A rotating disk with a vertical velocity disper-
sion characteristic of a spiral galaxy disk (10–
20 km s
1
) will have clouds superimposed in ve-
locity sp ace, and an overall kinematic spread of
tens of km s
1
. Strong Mgii absorption has been
found to arise along nearly all lines of within
40 kpc of normal galaxies (i.e. the covering fac-
tor is nearly unity w ithin that radius). The large
variety of kinematics evident in Mgii absorption
profiles is, in fact, consistent with a superposi-
tion of disk and radial infall (halo) motions, and
not with just one or the other. In addition to
these simple, toy models, ins ights can be gleaned
by passing lines of sight throu gh the structur es
in cosmological N–body/hydrod ynamic simula-
tions. In a few studies, metals have been added
7
= 200 km/s
rot
V= 200 km/s
infall
V
Radial Infall
Rotating Disk
-200 0 200 km/s
observer
observer
QSO
QSO
200 km/s0-200
-200 0 200
0
1
Fig. 5.— Illustrations of two simple kinematic models are shown in the top panel. To the left, the model is radial infall
of clouds to the center of a sphere, with constant velocity. The line of sight passes through five clouds, which leads to five
different absorption features (for a single transition) in the quasar spectrum. Two of the features are blueshifted relative
to the standard of rest of the absorbing galaxy, and the other three are redshifted. The absorption features from a radial
infall model can be spread over a velocity of 100–200 km s
1
, typical of the velocity dispersion of a galaxy halo. To the
right, a rotating disk model is illustrated. In this case all the “clouds” along the line of site have a component of motion
that is redshifted, and they tend to be clustered together in velocity space, with typical spread of 20-60 km s
1
. The lower
panel shows a sample of 0.4 < z < 1.4 Mgii absorption profiles observed with the Keck/HIRES spectrograph at R = 45, 000,
corresponding to a resolution of 6 km s
1
. The solid lines through these data are Voigt profile fits and the ticks drawn
above th e spectrum represent the cloud velocities. Some of these profiles are consistent with the kinematics of a rotating
disk, and others with radial infall kinematics. H owever, to explain the full ensemble of profiles a model combining these two
basic types of kinematics is needed.
8
uniformly throughout the simulation box and
photoionization m odels used to predict the ab -
sorption expected from different structures. T his
is especially important for establishing the kine-
matics that would be observed from the p rocess
of structure formation at high reds hifts.
3.2. Photoionization Models
Consider a cloud of material, modeled by a
plane parallel slab with a certain total column
density of Hydrogen, N(H) = N(Hi) + N(Hii),
and with a constant total number density n
H
=
n(Hi) + n(Hii) along the line of sight. The cloud
is also characterized by its metallicity, Z, which
is the ratio of Fe/H expressed relative to the so-
lar value, Z
, and by an abundance pattern (the
abundance ratios of all other elements to Fe).
The degree of ionization in the gas depends upon
the intensity and shape of the spectrum of ioniz-
ing radiation. T he intensity is characterized by
the ionization parameter, U = n
γ
/n
H
, which is
the ratio of the number density of photons at the
Lyman edge to the number density of Hydrogen
(n
H
= n
e
, where n
e
is th e total number density
of electrons). The larger the value of U, the more
ionized the gas. Collisional ionization can also be
an important process for some absorption sys-
tems with gas at high temperatures (hundreds of
thousands of degrees). Photoionization equilib-
rium models typically yield temperatur es of tens
of thousands of degrees.
Once the metallicity, abund ance pattern, ion-
ization parameter, and spectral shape are spec-
ified the equations of radiative transfer can be
solved to nd the column densities of all th e dif-
ferent ionization states of various chemical ele-
ments. Figure 6 illustrates, for N(Hi) = 10
16
and 10
19
cm
2
, the dependence of column densi-
ties of various transitions on the ionization pa-
rameter, U. For optically thin gas [N(Hi) <
10
17.2
cm
2
], the column density ratios of the
various metal transitions are not dependent on
the overall metallicity, i.e. the curves shift verti-
cally in proportion to Z. For optically thick gas,
10
12
14
16
10
12
14
16
10
12
14
16
-4.5 -4 -3.5 -3 -2.5 -2
10
12
14
16
-4.5 -4 -3.5 -3 -2.5 -2
10
12
14
16
-4.5 -4 -3.5 -3 -2.5 -2
10
12
14
16
Fig. 6.— Photoionization model predictions of the col-
umn densities of Mgii, Feii, and Civ as a function of
the ionization parameter (the ratio of ionizing photons to
the electron number density in the gas). The spectrum
incident on the cloud, represented by a constant den-
sity slab, is the “Haardt–Madau” spectrum (attenuated
spectrum due to integrated effect of quasars and young
galaxies). The predicted column densities are presented
in two series of mo dels with N(Hi) = 10
16
cm
2
and with
N(Hi) = 10
19
cm
2
, the optically t hin and optically thick
cases. For both, the metallicity is fixed at 10% of the so-
lar value. For the optically thin case, the column densities
scale with metallicity, i.e. the ratios remain constant, but
for the optically thick case the situation is more complex.
9
ionization structure d evelops, with an outer ion-
ized layer around a neutral core, and there is no
simple scaling relation with metallicity.
In practice, if we assume that a cloud has a
simple, single phase structure, the ratios of the
column densities can be used to infer the ioniza-
tion parameter, which relates to th e d en s ity of
the gas. However, the abundance pattern can
differ from the solar abundance pattern because
of differing degrees of depletion onto dust, or
because of different processing histories. Most
of the so–called α particle nuclei (such as Mg
and Si) are synthesized primarily by Type II su-
pernovae during the early history of a galaxy
when m ost massive stars form and q uickly evolve
to reach their end states. On the other hand,
the Fe–group elements are primarily produced
by Type Ia supernovae, and therefore build up
over a longer timescale. In the basic picture of
galaxy evolution, the halo stars are formed early,
have been en riched only by Type II supernova,
and therefore are α–element en hanced. Younger
disk stars have incorporated also the Type Ia
processed material and therefore have relatively
larger Fe–group abundances. Ideally, several dif-
ferent ionization states of the same chemical ele-
ment are observed so that there is no ambiguity
between the ionization parameter and the abun-
dance pattern, but this has generally not yet been
possible because of limited wavelength coverage
at high resolution.
Examples of the variation of column density
ratios with velocity in two absorption systems
are shown in Figures 7 and 8. In Figure 7,
N(Feii)/N (Mgii) varies by an order of magni-
tude over the four components in the z = 1.325
system toward the quasar Q0117+213. T his rep-
resents a variation of an order of magnitude in
the ionization parameter (10
4
< U < 10
3
), or
an order of magnitude variation in the abundance
pattern. Figure 8 is a very unusual system w ith
two clouds separated by only 20 km s
1
in veloc-
ity, one of which has a Silicon to Aluminum ra-
tio similar to the Milky Way ISM, and the other
which requires a significant enhancement of Alu-
minum.
0
.5
1
1.5
| | | | | |
-200 -100 0 100 200
0
.5
1
1.5
| | | | | |
Fig. 7.— HIRES/Keck Feii and Mgii absorption profiles
for the z = 1.325 system in the spectrum of the quasar
Q0117 + 213. The six clouds in this system show a range
of more than an order of magnitude in N(Feii)/N(Mgii),
given below each cloud in the lower panel. These varia-
tions could be due to cloud to cloud variations of ioniza-
tion parameter (density) or of abundance pattern within
the system.
4. Multiphase Conditions
The gaseous component of the th e Milky Way
and nearby galaxies have phase structure (i.e.
spatial locations with different densities and/or
temperatures). Examples are the disk/halo inter-
face (Galactic coronae) and th e cold, warm, and
hot phases of the interstellar medium. From pho-
toionization models, it is not usually possib le to
generate absorption that is simultaneously con-
sistent with all observed chemical transitions for
a given s ystem. For example, in single cloud Mgii
systems, Figure 6 (with N(Hi) = 10
16
cm
2
)
shows that if Feii is detected at a similar column
density to Mgii, the ionization parameter must
10
-200 -100 0 100 200
0
1
| |
0
1
| |
0
1
| |
0
1
| |
0
1
| |
Fig. 8.— An unusual Aluminum–rich cloud is apparent
in the z = 1.93 system toward the quasar Q1222+228, and
it is close in velocity sp ace to a normal (relative to Galactic
clouds) cloud which has detected Siii. Note the different
kinematic structure in the higher ionization transitions.
The excess of Alii and Aliii in the cloud at v = 9 km s
1
is best explained by an abun dance pattern variation, since
Siii and Alii are transitions with very similar ionization
states.
be small, and W
r
(Civ) cannot be large. Many
systems have Civ absorption w hich exceeds this
limit an d requires a higher ionization (lower d en -
sity) phase; generally, this phase must have struc-
ture over a large velocity range (a large “effec-
tive” Doppler parameter). The z = 0.93 system
toward the quasar PG 1206+459 is another case
that requires multiphase structure. The observed
Civ profile in Figure 4 is much too strong for this
absorption to arise in the same clouds that pro-
duce the Mgii, even if their ionization parameters
are pus hed to the largest values consistent with
the data.
5. Statistics, Evolution, and Interpreta-
tion
Future quasar absorption line studies will com-
bine insights gained from detailed analyses of in-
dividual systems with conclusions drawn from
the large statistical samples assembled over cos-
mic time. Evolution of the ensemble of absorp-
tion profiles generated by the universal collective
of intervening structures is a result of the com-
bined effects of numerous processes. These in-
clude growth of s tr ucture, star f ormation, mor-
phological evolution of galaxies, galaxy mergers,
and changes in the extragalactic background ra-
diation. Here, we summarize the best present
statistical data and likely interpretations for the
different classes of absorbers. The number of
lines per unit redsh ift for various populations of
absorbers is represented by a power law dN/dz
(1 + z)
γ
. For a universe with only the cosmo-
logical evolution due to expans ion, γ = 1.0 for
deceleration parameter q
0
= 0 and γ = 0.5 for
q
0
= 0.5.
5.1. Lyα Forest
The Lyα forest evolves away dramatically
from high to low redshift, as is strikingly clear
from the spectra of z 3 and z 1 quasars
in Figure 9. The evolution of the Lyα lines with
W
r
(Lyα) > 0.3
˚
A can be characterized by a dou-
11
PG1634+706
z=3.63
z=1.33
Q1422+2309
z=3
z=1
Fig. 9.— Illustration of stru cture evolution of intergalactic gas from high t o low redshift. The upper spectrum of a z = 3.6
quasar is a Keck/HIR ES observation, while the lower spectrum is a FOS/HST observations of a z = 1.3 quasar. Higher
redshift quasars show a much thicker forest of Lyα lines. Slices through N–body/hydrodynamic simulation results at the
two epochs z = 3 and z = 1 are shown in the right–hand panels. Three contour levels are shown: 10
11
cm
2
(dotted lines),
10
12
cm
2
(solid lines) and 10
13
cm
2
(thick solid lines). Evolution pro ceeds so that the voids become more empty so that
even the low column d ensity material is found in lamentary stru ctures at low redshifts.
ble power law with γ 2 for 1.8 < z < 4.5
and γ 0.2 for z < 1.8. Help in understand-
ing the physical picture has come from sophis-
ticated N–body/hydrodynamic simulations that
incorporate the gas physics and consider cosmo-
logical expansion of the simulation box. The dy-
namical evolution of the Hi gas can be described
as outflow from the centers of voids to their sur-
rounding shells, and flows along these sheets to-
ward their intersections where the densest struc-
tures form. This picture is consistent with ob-
servational determinations of the “sizes” of Lyα
structures. It is difficu lt to obtain direct mea-
surements of sizes except in some special cases
to use “double lines of sight”, close quasar pairs,
either physical or apparent due to gravitational
lensing. If the spectra of the two quasars both
have a Lyα absorption line at the same wave-
length that implies a “structure” which covers
both lines of sight. From these studies, it is found
that “structures” are at least hundreds of kpc in
extent.
At redshifts z = 5 to z = 2 dN/dz for Lyα for-
est absorption is quite large, but it is declining
very rapidly over that range. This dr amatic evo-
lution in the number of forest clouds is mostly
due to the expansion of the universe, with a
modest contribution from structure growth. At
z < 2, the extragalactic background radiation
field is falling, and Lyα structures are becom-
ing more neutral. Therefore, the more numerous,
smaller N(H) structures are observed at a larger
N(Hi) and this will counteract the effect of ex-
pansion, thus slowing the decline of the forest.
The high redshift Lyα forest was once thought
12
to be primordial m aterial, b ut in fact it is ob-
served to have a metallicity of 0.1% solar, even
at z = 3. For N(Hi) < 10
14
cm
2
, the expected
N(Civ) would be below the detection thresholds
of current ob s ervations, s o truly pristine mate-
rial still eludes us. Perhaps it does not exist.
To spread metals all through the intergalactic
medium may have required a “pre–galactic” pop-
ulation of stars at z > 10 that polluted all of
intergalactic space.
5.2. Lyman Limit and Metal Line Sys-
tems
The dN/dz of Lyman limit sys tems is consis-
tent with that of str ong Mgii absorbers [with
W
r
(Mgii) > 0.3
˚
A] over the redshift ran ge for
which both have been observed, 0.4 < z < 2.2.
For W
r
(Mgii) > 0.3
˚
A, γ = 1.0 ± 0.1, consistent
with no evolution. For even stronger Mgii sys-
tems [W
r
(Mgii) > 1
˚
A], dN/dz increases more
dramatically with z, with γ = 2.3 ± 1.0.
The number of Mgii systems (equivalent width
distribution) continues to increase down to the
sensitivity of the best surveys, W
r
(Mgii) > 0.02
˚
A,
such that dN/dz = 2.7 ± 0.15 at z 1. The
“weak” Mgii absorbers are th erefore more com-
mon than the strong systems [W
r
(Mgii) > 0.3
˚
A]
known to be associated with luminous galaxies.
Unlike the str ong Mgii absorbers, the weak Mgii
absorbers are sub-Lyman limit systems (they do
not have Lyman limit breaks), and no galaxies
have been identified at the redshift of absorp-
tion. Yet, photoionization models indicate th at
the metallicities of these weak absorbers are at
least 10% of the solar value, and in some cases
comparable to solar. They are a varied popula-
tion: some have relatively strong Feii while oth-
ers have no Feii detected, and some have strong
Civ that requires a separate phase while others
have no Civ detected. Those with strong Feii
are constrained to be smaller than 10 pc (the
ionization parameter must be small and n
e
large
as can be seen in Figure 6). Also, since Fe is
produ ced primarily by Type Ia supernovae they
must be enriched by a relatively old stellar pop-
ulation. Those with weaker or undetected Feii
could be larger (kpcs or tens of k pcs) and pos-
sibly enriched by Type II supernovae. Candi-
date environments that could be traced by weak
Mgii absorption are: remnants of pre–galactic
star clusters formed in mini–halos at z > 10, su-
per star clusters formed in interactions, tidally
stripped material, low surface brightness galax-
ies, and ejected or infalling clouds (analogous to
the Milky Way high velocity clouds).
The evolution of dN/dz for Civ abs orbers can
be studied in the optical for high redshifts. For
W (Civ) > 0.4
˚
A and z > 1.2, the number de-
creases with increasing z, as γ = 2.4 ± 0.8. In
this same interval, the number of Lyman limit
systems is still increasing with redshift, with
γ = 1.5 ± 0.4. This implies that the dramatic
evolution in the number of Civ sys tems is either
due to a change in metallicity or a change in ion-
ization state. T he dN /dz for Civ systems peaks
at intermediate z and declines, consistent with
no evolution until the present. Combining opti-
cal and UV data, Civ and Mgii have been com-
pared at 0.4 < z < 2.2. The fraction of systems
with large W
r
(Civ)/W
r
(Mgii) decreases rapidly
with decreasing redshift; there is a shift toward
“lower ionization systems”.
It is important to consider that the Hi, Mgii,
and Civ absorption do not always arise in the
same phase. It is possible that the Civ in many
z 1 Mgii absorption systems arises in a phase
similar to the Galactic coronae. If the origin of
this phase is related to star–forming processes in
the disk, then it might be expected to diminish
below z = 1.2 since the peak star formation rate
is passed.
Another important trend is the fact that th e
very strongest Mgii absorbers evolve away from
z = 2 until the present. If we study the kinematic
structure of these objects, we find that they com-
monly have a “double” structure, with two sepa-
rate kinematic regions in the Mgii profile. These
objects also have strong Civ which also has sepa-
13
rate components around the two Mgii regions in
the “double” structure. The Civ does not arise
primarily in the individual Mgii clouds, nor is it
in a smooth, “common halo” stru ctur e that ex-
tends in velocity space around the entire Mgii
profile. As more data are collected on the kine-
matic structure of various transitions in these
“double” systems, it will be interesting to con-
sider the hypothesis that galaxy pairs in the p ro-
cess of merger are responsible. Th e number of
these is thought to have been dramatically larger
in the past.
5.3. Damped Lyα Systems
The N (Hi) > 10
20.3
cm
2
systems are of par-
ticular interest because it is possible to observe
many different chemical elements (such as Zn,
Cr, Fe, Mn, and Ni) in these objects back to
high redshift. Metallicities and abundance pat-
terns can be studied and compared to those of
old stellar populations in the Milky Way. Back
to z = 3, the metallicity in DLAs, as measured
by the undepleted element Zinc, is about 10% of
the solar value, but it may decline at z > 3. The
identity of sites responsible for DLAs at high z re-
mains controversial, b ut they do contain most of
the neutral Hydrogen in the universe, from which
most of its stars form. Th e kinematic structure
of the absorption profiles of neutral and low ion-
ization species is consistent with the rotation of
a thick disk, so that it is possible that these are
the z = 3 progenitors of normal spiral galaxies.
However, this signature is not unique. It could
also be the consequence of directed infall in an
hierarchical structure formation scenario. The
higher ionization species show complex kinemat-
ics which vary in relation to those of the lower
ionization gas; in some s ystems they appear to
trace relatively similar structure, and in others
there are clearly several different phases.
At low redshift, many of the galaxies that are
responsible for the DLA absorption can be di-
rectly identified. These galaxies are a heteroge-
neous population. They are not just the most lu-
minous galaxies, but includ e dwarf and low sur-
face brightness galaxies, and even cases where
no galaxy h as been identified to sensitive limits.
Damped Lyα absorption does n ot trace the most
luminous objects, but rather it tr aces the largest
neutral gas reservoirs. An additional selection
effect may be important. The most dust–rich
galaxies that have the potential to produce DLA
absorption could produce enough extinction th at
their background quasars will not be included in
quasar surveys. In this way, the population of
DLAs that are actually observed could be signif-
icantly biased against dusty galaxy hosts.
6. Future Prospects
The next decade will see the synthesis of the
various techniques for the study of galaxy evo-
lution, through their stars and through their
gas. Higher resolution quasar spectra will be
obtained in the ultraviolet (with the Space Tele-
scope Imaging Spectrograph (STIS) and with the
Cosmic Origins Spectrograph (COS) on the HST,
and later, hopefully, with a larger UV space tele-
scope). It will then be possible to conduct a sys-
tematic analysis of the relationship s between the
different ionization species that trace the differ-
ent phases of gas in 0.4 < z < 1.5 galaxies. In
this redshift regime, comparisons to the detailed
morphological s tr ucture and orientations of the
absorbing galaxies is possible from HST images.
Invaluable ins ights into the origin of quasar
absorption lines have been gleaned from absorp-
tion studies of nearby galaxies, for which it is
possible to directly observe the processes that are
involved. Making more obs ervations of this type
will be possible by discoveries of bright quasars
that fall behind nearby galaxies. The discoveries
of quasars in large sur veys will also include mul-
tiple lines of sight behind distant absorption line
systems which can be used to produce 3–D maps
of the structures.
The interstellar medium of the Milky Way
shows structur e on sub–pc scales, and absorp-
14
tion features can only be resolved with resolution
< 1 km s
1
. Such a resolution will soon be avail-
able on 8m–class telescopes. This is important
for separating blends and for looking for metallic-
ity, ionization, and abundance pattern gradients
along the line of sight.
The key low ionization transitions of Mgii and
Feii are shifted into the near–IR region of the
spectrum for z > 2.5. Very soon, near–IR quasar
spectra will be obtained at relatively high resolu-
tion ( 20 km s
1
). Also, IR–imaging, n arrow–
band tech niques, and multi–object spectroscopy
in the near–IR sh ould provide much more infor-
mation about absorbing galaxies at higher red-
shifts. This will extend evolutionary studies back
to an epoch at which formation processes may be
contributing significantly to evolution.
7. Bibliography
Articles in review journals:
Rauch M 1998 The Lyman Alpha Forest in the
Spectra of QSOs ARAA 36 267
Churchill C W an d Charlton J C 2000 Mgii Ab-
sorbers: A Review PASP in press
Conference proceedings:
Blades J C, Turnshek D A, and Norman C 1988
QSO Absorption Lines: Probing the Universe,
Proceedings of the QSO Absorption Line Meet-
ing, Baltimore 1987 (Cambridge: Cambridge
University Press)
Meylan G 1995 QSO Absorption Lines: Proceed-
ings of the ESO Workshop, Munich 1994 (Berlin:
Springer)
Petitjean P and Charlot S 1997 Structure and
Evolution of the Intergalactic Medium from QSO
Absorption Lines (Paris: Editions Fronti´eres)
15
... This is done by adopting a forward selection model [Pahikkala et al., 2010], based on repeated kNN experiments on random subsamples of the available dataset, in order to generate a tree of features, from which to select the best performing branch. Those features have been then physically interpreted with respect to the typical spectral emission lines for quasars [Charlton and Churchill, 2000]. The final purpose of establishing these two different and, from a certain point of view, alternative methods, is to give to the community a good and affordable way to estimate photometric redshifts for a huge number of sources. ...
Thesis
The problem of photometric redshift estimation is a major subject in astronomy, since the need of estimating distances for a huge number of sources, as required by the data deluge of the recent years. The ability to estimate redshifts through spectroscopy does not scale with this avalanche of data. Photometric redshifts provide the required redshift estimates at the cost of some precision. The success of several forthcoming missions is highly dependent on the availability of photometric redshifts. The purpose of this thesis is to provide innovative methods for photometric redshift estimation. Two models are proposed. The first is fully-automatized, based on the combination of a convolutional neural network with a mixture density network, to predict probabilistic multimodal redshifts directly from images. The second model is features-based, performing a massive combination of photometric parameters to apply a forward selection in a huge feature space. The proposed models perform very efficiently compared to some of the most common models used in the literature. An important part of the work is dedicated to the correct estimation of the errors and prediction quality. The proposed models are very general and can be applied to different topics in astronomy and beyond.
ResearchGate has not been able to resolve any references for this publication.