In 1963, Maarten Schmidt (Schmidt 1963) reported the discovery of a remarkable stellar-like object, 3C 273, which appeared at a ``high redshift'' (z = 0.16). This object was clearly not a star - it was detected at radio wavelengths and its spectrum showed very broad emission lines. This quasi-stellar object (QSO) would define a new class of objects - quasars. QSOs are now known to be the highest luminosity end of a much larger group of objects, active galaxies, which populate the universe. In terms of structure, content and morphology, active galaxies are no different from inactive, or normal, galaxies and thus cannot be distinguished from simple optical images as shown in Figure 1.1. (QSOs can be distinguished since light from the central core far outshines the galaxy, and therefore generally appear as point sources.)

Spectroscopically, active and normal galaxies are very different. The spectrum of a normal galaxy consists of the sum of all the starlight and reprocessed starlight (e.g., by interaction with dust). Figure 1.2 shows typical spectra of normal elliptical and spiral galaxies. They consist of a continuum with superposed absorption lines and, in some cases, narrow emission lines (<300 km/s) like H-beta. The shape of the continuum, and fingerprint of absorption lines depends on the distribution of stars contained within the galaxy. Elliptical galaxies have spectra reflective of late-type stars and old stellar populations. Spiral galaxies tend to have spectra of early-type stars. The detectable portion of the spectrum generally extends from the near ultraviolet (dominated by blackbody radiation from high mass stars) to the sub-millimeter (dominated by reprocessed starlight from dust), with a peak in the near infrared. For example, the luminosity of the Milky Way, a normal spiral galaxy, rises from 3x10^38 erg/s at radio wavelengths, peaks at 3x10^43 erg/s in the optical regime, and drops back down to 10^39 erg/s at X-ray energies (Zombeck 1990).

On the other hand, the rest-frame spectrum of an active galaxy, especially a QSO, is observable over several decades of energy, from the radio to hard X-rays, with a nearly constant luminosity per unit decade (L ~ 10^{42-46} erg/s). In the radio, synchrotron emission arises from powerful collimated jets in 10\% of active galaxies and QSOs. At very large distances (on megaparsec scales) these jets crash into the intergalactic medium which then produces roughly isotropic, self-absorbed synchrotron emission. In the sub-millimeter and infrared, emission is likely dominated by dust. At optical, ultraviolet and X-ray wavelengths, continuum emission arises from a compact point-like (i.e., as yet unresolved) core (probably an accretion disk around a supermassive black hole). Surrounding the compact continuum region is a more extended region producing emission lines with widths up to ~10,000 km/s. The size of this broad emission line region (BELR) ranges from light-days to light-months, as measured from reverberation mapping (e.g., Kaspi et al 2000). The size of this region is also dependent on the specific line - emission from high ionization species (e.g., C IV, N V) tend to originate from closer to the continuum source than low ionization (e.g., Mg II) and high excitation (e.g., H beta) species. At yet larger distances from the continuum region/BELR is a region that produces very narrow forbidden lines with velocity dispersions on par with the stellar velocity field of the host galaxy. In the next section, I report on a model, the accretion-disk/wind scenario, that elegantly explains the optical, UV, and X-ray continuum and broad emission line regions (that is, the ``central engine'') of active galaxies.

Section 2: The Accretion-Disk/Wind Scenario

The accretion-disk/wind scenario describing QSOs (see Figure 1.3) holds that a disk of matter orbits a supermassive black hole. This disk extends from the last stable orbit out to 10^{4-5} gravitational radii (R_g = GM/c^2 ~ M_8 AU, where M_8 is the mass of the black hole in units of 10^8 Msun). For a Schwarzchild black hole, the last stable orbit occurs at 6R_g. For a Kerr black hole, the radius of the last stable orbit depends on the alignment between the angular momentum of the accretion disk and the spin of the black hole. In general, of course, fuel from kiloparsec-scale distances comes in at an arbitrary angle; there is no evidence of alignment between the accretion disk and, say, the disk of the host galaxy. Bardeen & Petterson (1975) showed that the Lense-Thirring effect, or ``inertial--frame dragging,'' from the black hole causes the accretion disk to align itself with spin of the black hole with the transition occurring between 10^2 and 10^4 R_g. For retrograde orbits, the radius of last stable orbit around a maximally rotating black hole is 9R_g. For prograde orbits, it is R_g. For the purpose of this work, it is only important to make the distinction insofar as the formation and collimation of radio jets may be linked to the spin of the black hole.

Through torques arising from viscous forces, magnetic fields, instabilities, mass loss, or a combination thereof, matter in the disk sheds angular momentum and spirals down past the last stable orbit of the black hole. Heating causes the disk to swell slightly (the thickness-to-radius ratio remains much less that unity) via gas pressure and the release of gravitational energy yields a source of radiation pressure which lifts material off the disk. In the absence of magnetic fields, the matter maintains its angular momentum - having, therefore, both a rotational and a vertical velocity. The rotational velocity decreases with radius as v_theta ~ 1/r. In the presence of a magnetic field, the matter co-rotates with and travels along field lines. In this case, the angular velocity, v_theta/r, remains constant.

Thus, the matter spirals off the disk (i.e., in a helix, with a vertical component from local-disk radiation pressure and a rotational component either from angular momentum conservation or magnetic field line co-rotation) until it feels radiation pressure from the inner accretion disk (~6-60 R_g) which is the source of (thermal) ultraviolet and soft X-ray continuum radiation. Through line pressure, the material is driven radially away. A rotation velocity component remains, however, falling off inversely with radius (from angular momentum conservation), or increasing linearly with radius (if the material co-rotates with the disk in the presence of magnetic fields). The velocity of the wind is capped by the terminal velocity, which is on the order of 10^4 km/s and scales with the luminosity of the disk (v ~ L^{1/2}). Detailed photoionization and hydrodynamic (Murray et al 1995; Proga, Stone, & Kallman 2000; hereafter MCGV and PSK) simulations have demonstrated that this wind can explain the shapes of broad emission lines. They have also shown that Kelvin-Helmholtz shearing instabilities result in discrete clumps of gas which hug the wind.

In a broader context, it is important to understand how the wind is formed and how its energetics are affected by the fundamental properties of the central engine - the mass and spin of the black hole, and the fuelling rate of the accretion disk (from the environment), and the mass consumption rate (or accretion rate). The wind may be an important mechanism by which angular momentum is transferred away from the accretion disk (allowing matter to accrete). In the accretion-disk/wind scenario described above, it has been demonstrated that the dynamics and physical conditions of the wind (e.g., level of ionization) play an integral role in the formation of the broad emission lines. Since broad emission lines are observed in almost all QSOs, it is also important to understand what can be learned about the central engine from them, and how this should be done. Even after decades of study, the source of emission lines is still uncertain. Thus, placing constraints on the type of structures that can produce them is interesting.

Section 3: Origins of Quasar Absorption Lines

In 1965, two years after the discovery of 3C 273, two letters appeared in volume 142 of the Astrophysical Journal which founded the field of quasar absorption lines. The first paper, Bahcall & Salpeter (1967), predicted that ubiquitous and clumpy gas parcels along the line of sight toward a distant background QSO would produce discrete Lyman-alpha absorption features at observed wavelengths of [1+z(abs)]x1215.67 A, where the absorber redshift, z(abs), lies between zero and the QSO redshift. A few years later, this would be observationally discovered and termed the ``Lyman-alpha forest'' (Lynds 1971). The other paper, Gunn & Peterson (1965), noted that if the gas between the observer and the QSO were not clumpy (e.g., at redshifts before the reionization of the universe), one would see a general suppression of the QSO's continuum radiation at wavelengths shorter than the Lyman-alpha emission line. The Gunn-Peterson effect, as this general suppression is termed, has been observed for both He II Lyman-alpha (e.g., Smette et al 2002, Heap et al 2000), and H I Lyman-alpha (e.g., Becker et al 2001, Fang et al 1998, Giallongo et al. 1994).

After 36 years of research, it is now known that quasar absorption lines arise from a very eclectic variety of objects (Weymann et al. 1979, Young, Sargent, & Boksenburg 1982, Foltz et al. 1986, Anderson et al. 1987, Lanzetta, Turnshek, & Wolfe 1987, Sargent, Boksenburg, & Steidel 1988, Bahcall et al. 1993, Jannuzi et al. 1998). The light from a distant QSO passes through the Galaxy, high velocity clouds (HVCs), other galaxies, the filamentary structure of the intergalactic medium, the cluster gas of the QSO, the host galaxy of the QSO, and high column density gas near the QSO central engine. All of these can produce a variety of narrow absorption lines (from a range of ionization species) at their respective redshifts.

While it has always been acknowledged that at least some fraction of absorption lines are formed from gas near the QSO central engine (hereafter, intrinsic absorption lines), a long-standing problem has been the separation of these lines from those due to intervening gas. Intrinsic absorption lines come in two varieties - with broad and narrow line-of-sight velocity dispersion. Broad absorption lines (BALs) have velocity widths on the order of 10^4 km/s (Turnshek et al. 1988, Weymann et al. 1991). These are explained as radiatively accelerated outflows from the accretion disk and are now accepted as a form of intrinsic absorption. Narrow absorption lines (NALs) are not as easily divided between intervening and intrinsic material. NALs have velocity widths that are sufficiently narrow to separate resonant doublets (e.g., Delta v = 498 km/s for C IV 1548.202,1550.770; Hamann & Ferland 1999).

There are two schools by which progress toward understanding the origins, and physical conditions of intrinsic NALs is made. Under one school, large samples of intrinsic absorbers are found by statistically identifying excesses of absorption lines over what is expected from the random population in space in some redshift path. This school has the ability to look at correlations of QSO properties with gross properties of the absorbers (e.g., equivalent width). The drawback of this school, is that it is never actually known which absorbers are truly intrinsic. As such, any apparent lack of correlations is to be taken lightly since there is a dilution of trends due to ``contamination'' by intervening absorption lines. By the same argument, statistically significant trends should be examined closely to ensure that intervening absorbers are not responsible.

In the other school, specific absorbers are shown to be intrinsic by demonstrating that the distance between the absorber and the QSO central engine is small (<1 kpc); that is, close to the central engine. This school has the advantage of being able to make detailed studies of the absorption profiles to glean the physical conditions (e.g., ionization parameters, densities) and geometry of the gas. This school is generally more observationally expensive as it either requires high resolution spectroscopy to look at resolved profile shapes, or several epochs of observation. Barlow & Sargent (1997) and Hamann et al. (1997a, b) list several characteristics which tend to separate intrinsic an intervening systems. The two most robust and easily observable features are time variability of profiles and partial coverage of the background source (the QSO itself).

Section 4: Unanswered Questions

Several questions still remain about intrinsic narrow absorption lines. What is their rate of incidence? Does NAL gas potentially exist in all QSOs? Where does that absorbing gas come from? Is it related to the gas in other regions (e.g., the broad emission line region)? Where is it in relation to the other regions? Is there only a single population of intrinsic NALs or are there different varieties? Is there a relationship between the properties of intrinsic NALs (e.g., presence, strength, type if there are separate populations) and the fundamental properties (e.g., black hole mass, mass accretion rate) or observed properties (e.g., luminosity, spectral shape, emission line shape) of the host QSO? Is it possible to understand these in the context of models for QSOs or can any of these models be dismissed? Answering these questions is the motivation of this research.

In the next section (§ 5), I will develop the formalism for the production of intrinsic narrow absorption lines and the measurement of partial coverage fractions. I will describe how partial coverage and time variability of profiles imply an intrinsic origin for the absorbing gas. The bulk of the material contained in the following chapters appear in The Astrophysical Journal and The Astronomical Journal. Where relevant, I provide a parenthetical citation. In chapter 2 (Ganguly et al. 1999), I will refine the technique of measuring partial coverage to separately include partial coverage of the compact continuum source (presumably the inner accretion disk) and the more extended broad emission line region (explicitly allowing me to address the question as to where the absorption line gas lies in relation to the BELR). I will apply this refined technique to a small sample of Keck/HIRES spectra and discuss the effects that instruments have on measuring partial coverage fractions. In chapter 3, I will present the first results from a low-spectral resolution snapshot program with the Hobby-Eberly telescope to search for variable absorption lines in high-redshift quasars [z(em)>2.5]. Also included in this chapter is the serendipitous discovery of a possibly intrinsic absorption complex toward the quasar CSO 118 (Ganguly, Charlton, & Eracleous 2001). In chapter 4, I will use a multivariate analysis on a sample QSOs observed with HST/FOS to explore the relationship between intrinsic NAL gas and the host QSO (Ganguly et al. 2001). In chapter 5, I will show a detailed case study of the intrinsic N V absorption complex toward RX J1230.8+0115. Finally, I will summarize what is now known about intrinsic NAL gas and synthesize this into a coherent model based on the disk-wind scenario to explain the properties of intrinsic NALs in chapter 6.

Section 5: Identification of Intrinsic Lines

The identification of a specific absorber as intrinsic to a QSO entails showing that the gas must lie close to the central engine. While it possible to constrain the absorber-source distance through detailed photoionization modelling, there are two observational ``smoking guns'' that can be readily employed in searches for intrinsic absorbers. The first feature is partial coverage of the engine by the absorbing gas. The second is rapid time variability.

Section 5.1: Formalism of Intrinsic Absorption Profiles & Partial Coverage

Clouds that lie near the background source of photons will not completely occult the source if their size is smaller than the size of the source. If the clouds lie close to this source, then light can also be scattered around them and into the observer's line of sight. These can be distinguished using spectropolarimetry since the scattered light will have a characteristically high degree of polarization. Distinguishing between geometry and scattered light is beyond the scope of this dissertation; it is only important to note here that both scenarios require that the absorbing clouds lie ``close'' to the background source. In either case, there is a fraction of sightlines which effectively reach the observer that are not occulted by the absorbing cloud. Let this fraction be 1-C. Then the flux that reaches the observer at a wavelength {$\lambda$} can be expressed in terms of an occulted and a non-occulted part:

Another form of broadening that is always present is natural or atomic broadening which is described by a Lorentzian function like the amplitude of a driven harmonic oscillator. Physically, this broadening is due to the finite lifetime of the transition (characterized by {$\Gamma^{-1}$) which creates an uncertainty in the energy of the transition. The lifetimes are on the order of 10~ns, translating to a characteristic broadening width of {$10^{-4}$~\AA}. This broadening is only important under very dense conditions where photons incident on a cloud are scattered into an energy where absorption becomes probable.

Formally, the total line-broadening function is the convolution of individual broadening mechanisms. In cases where absorption is done by clouds, the only mechanisms that are important are thermal broadening, turbulence, and natural broadening. Thus, the line-broadening function in QSO absorption line research is the convolution of a Gaussian and a Lorentzian -- a Voigt profile. A simple way to numerically compute the Voigt function was given by {\citet{humlicek79}} as the real part of the complex probability function, \begin{equation} u(x,y) = Re \left [ e^{-z^2} \left ( 1 + \int_{0}^{z} e^{-t^2} dt \right ) \right ], z = x + iy, \end{equation} where {$x$} and {$y$} are defined as follows: \begin{equation} x = {{\lambda - \lambdarest} \over {\lambdadop}}, y = {{\lambdarest^2 \Gamma} \over {4 \pi c \lambdadop}} \end{equation}

For a collection of {$m$} clouds, each having a different coverage fraction {$C_{\mathrm{i}}$} ({i$=1 \ldots m$}), equation~\ref{eq:fluxobs} changes to \begin{equation} I_{\mathrm{obs}} = \prod_{i=1}^{m} [(1-C_{\mathrm{i}}) + C_{\mathrm{i}} e^{-\tau_{\lambda,i}}] \times I_{\mathrm{em}}, \label{eq:cfmclds} \end{equation} essentially providing an attenuation factor for each cloud. One subtlety occurs if the cloud has an ionization structure. In this case, for each cloud, there can be a different coverage fraction for each ionization species. In this sense, it is often best to allow the coverage fraction to also be a function of wavelength (that is, a function of the particular transition one is considering).

Another scenario that can create absorption features in a QSO spectrum is absorption by an outflowing wind. In this case, there is a small difference in this formalism. As far as absorbing photons are concerned, a wind can be treated as a single absorber which has velocity structure (as opposed to random motions centered as a given velocity or redshift). Let the number of particles in the wind with velocities between {$v$} and {$v + dv$} be {$N(v) dv$} (where {$v$} is the line of sight velocity). Then the optical depth of the wind at wavelength {$\lambda$} due to a transition with rest frame wavelength {$\lambdarest$} is {\citep{ss91}}: \begin{equation} \tau_{\lambda} = {{{\pi} e^2} \over {m_e c}} {N(v) g f \lambdarest}, \label{eq:optdepthwind} \end{equation} With a coverage fraction that is also a function of velocity (and therefore wavelength through the Doppler shift), the observed profile is given by: \begin{equation} I_{\mathrm{obs}} = [1-C_f(\lambda) + C_f(\lambda) e^{-\tau_{\lambda}}] \times I_{\mathrm{em}}, \end{equation} In chapter~\ref{sec:pcrefine}, I discuss the effect of the instrument response on the absorption profile and the measurement of the coverage fraction.

\subsection{Time Variability} \label{sec:timevar}

Variability of a profile (strength, shape) over time can arise from one of two reasons: (1) motion of the absorbing gas across the line of sight (hereafter, bulk motion); or (2) changes in the ionization state of the gas (hereafter, ionization/recombination). Time variability of spectral profiles can be demonstrated in either of two ways. If the variability is sufficiently extreme, then the equivalent width of the profiles should change. However, even if variability is more subtle, it can still be quantitatively assessed by comparing the changes in the kinematic distribution and column densities of components. Fig.~1.4 demonstrates how one can observationally distinguish between the two forms of variability. If bulk motion is the dominant source of variability, then both low and high ionization lines will change together (both get weaker or both get smaller). If recombination/ionization dominates the variability, the different ionization species will change inversely (one will get weaker while the other becomes stronger). Thus repeated observations can determine the form and timescale of variability. {\citet{ham95,ham97b}} outline how the timescale of variability can be used to infer the distance between the absorber and the central engine.

\begin{figure} \begin{center} \rotatebox{270}{\scalebox{0.5}{\includegraphics{./Intro_Figs/galaxycomp.eps}}} \end{center} {\caption[Comparison of an Active and Inactive Galaxy]{{\bf Comparison of an Active and Inactive Galaxy:} The above panels show {$11' \times 11'$} images of two spiral galaxies taken from the Digitized Sky Survey. The left panel is an image of NGC 4321, an inactive galaxy in Virgo. The right panel is an image of NGC 1566, an active galaxy (Seyfert 1) in Dorado.}} \label{fig:galaxycomp} \end{figure}

\clearpage

\begin{figure} \begin{center} \vglue -1.0in \rotatebox{0}{\scalebox{0.7}{\includegraphics{./Intro_Figs/galaxyspec.eps}}} \end{center} \vglue -0.7in {\caption[Spectra of Inactive Galaxies]{ {\bf Spectra of Inactive Galaxies:} Template spectra of inactive elliptical (left panels, denoted `E') and spiral (right panels, denoted `S') galaxies \citep{bica88}. The number refers to the placement of the galaxy in the Hubble sequence, which is based on the ellipticity and bulge dominance. The spectra of elliptical galaxies are dominated by late-type and evolved stars. Spiral galaxies are dominated by young stellar populations. For the spiral galaxies, only the contribution from the bulge is shown. The vertical lines demark {Ca\,{\sc ii}$\lambda\lambda3934,3968$}, {H$\beta4841$}, {Mg\,$\lambda5183$}, {Na\,{\sc i}$\lambda\lambda5890,5896$}, {H$\alpha6563$}, A and B bands (atmospheric absorption at 7594~\AA and 6867~\AA), and {Ca\,{\sc ii}$\lambda8542$}.}} \label{fig:lazygalaxy} \end{figure}

\begin{figure} \begin{center} %\vglue -5.0in \rotatebox{0}{\scalebox{1.0}{\includegraphics{./Intro_Figs/diskwind.eps}}} \end{center} %\vglue -0.5in {\caption[The Accretion-Disk/Wind Scenario]{ {\bf The Accretion-Disk/Wind Scenario:} This is a schematic illustration of the disk--wind scenario {\citep{mur95}} which offered an explanation of the BAL phenomenon. The central region of the accretion disk (dark shade) is the source of continuum UV and X--ray photons, while the inner portion of the wind (light shade) is the source of UV high--ionization line emission (e.g., {\CIV}, or {\NV}) and {\Lya}. In this picture, NAL gas is viewed as a clumpy medium that ``hugs'' the wind relatively far from the inner accretion disk.}} \label{fig:diskwind} \end{figure}

\begin{figure} \begin{center} \rotatebox{0}{\scalebox{0.5}{\includegraphics{./Intro_Figs/variability.eps}}} \end{center} \vglue -0.8in {\caption[Observational Signature of Variability]{ {\bf Observational Signature of Variability:} In the middle column, I show a velocity-aligned plot of a 4-cloud model system with the following parameters: {$\log U = -1.0$} for all clouds (see chapter~\ref{sec:rxj1230} for a definition of the ionization parameter, $U$), {$\log N_{\mathrm{H}} = 19.5, 19.5, 19.0, 19.5$} and {$C_{\mathrm{f}} = 0.6, 1.0, 1.0, 0.6$}. The four transitions shown are {\Lya}, {\SiIV$\lambda1393$}, {\CIV$\lambda\lambda1548,1550$} (centered on {$\lambda1548$}), {\NV$\lambda\lambda1238,1242$} (centered on {$\lambda1238$}). In the left column, I show what is expected if bulk motion (clouds streaming across the line of sight) causes variability. Here I decreased the coverage fraction by 0.3 and the column densities by 0.5 dex. In the right column, I show the result of decreasing the ionization parameter to {$\log U = -2.0$} -- what is expected by recombination if a change in the ionization state causes variability.}} \label{fig:variability} \end{figure}