MONITORING THE VARIABILITY OF INTRINSIC ABSORPTION LINES IN QUASAR SPECTRA^*,^†,^‡

We have selected quasars with intrinsic NALs based on the partial coverage indicator: the dilution of absorption troughs by unocculted light (e.g., Wampler et al. 1995; Barlow & Sargent 1997) from the background sources. The basic test relies on the optical depth ratio of resonant, rest-frame UV doublets of lithium-like species (e.g., C iv, N v). When this ratio is inconsistent with 2:1, as dictated by atomic physics (for fully resolved unsaturated absorption profiles), the deviation is interpreted as dilution of the absorption troughs by an unocculted portion of the background source. The optical depth ratio provides a means of computing the fraction of background light occulted by the absorber (the "coverage fraction"): C_f = (R₁ − 1)²/(R₂ − 2R₁ + 1), where R₁ and R₂ are the continuum-normalized intensities of the weaker and stronger components of the doublet. Recently, Misawa et al. (2007a) surveyed 37 quasars at redshift 2 < z < 4 and found that at least 43% had one or more intrinsic NALs as evidenced by partial coverage in the C iv, N v, and/or Si iv doublets, which implies that a substantial fraction of the solid angle around a substantial fraction of quasars is covered by material related to the quasar. It is more difficult to determine with certainty the origin of this material, more specifically whether it originates in the host galaxy, or whether it is a part of the quasar wind itself.

Our systematic study is based on long-term monitoring observations of 11 quasars (at z_em ∼ 2.00–3.08) with at least one intrinsic NAL or mini-BAL. We select sample quasars from three sources; (1) quasars with intrinsic NALs that were identified in Misawa et al. (2007a), (2) quasars with mini-BALs that we identified from the VLT/UVES archive sample of Narayanan et al. (2007), and (3) quasars with mini-BALs already published in the literature (Hamann et al. 1995, 1997a; Dobrzycki et al. 1999). These intrinsic NALs and mini-BALs were identified through their partial coverage (as described above) or their broad absorption profile. Thus, our sample quasars are all bright enough for high dispersion spectroscopy, and classified into a category of very luminous quasars. Another essential selection criterion is that at least two previous high-signal-to-noise ratio (S/N), high-resolution (R = 35, 000–45,000) spectra already exist, taken at different epochs with either the Keck/HIRES, the VLT/UVES and/or the Subaru/HDS. The sample quasars that satisfy both criteria are summarized in Table 1.

Four of our quasars come from a quasar sample that was originally selected and observed for measuring the deuterium-to-hydrogen abundance ratio (D/H) in the Lyα forest (e.g., Tytler et al. 1996). The typical value of D/H is so small, 2–4 × 10⁻⁵ (O'Meara et al. 2001 and references therein), that we can detect only D i lines corresponding to H i lines with large column densities, log ($N_{ {\rm H}\,\scriptsize{I}}$/cm⁻²) ⩾ 16.5. Therefore the survey included 40 quasars, in which either damped Lyα systems or Lyman limit systems were detected. This target selection method does not directly bias our sample with respect to the properties of any intrinsic absorption-line systems that these spectra may contain. The observations were carried out by a group led by David Tytler, using Keck/HIRES with a 114 slit resulting in a velocity resolution of ∼8 km s⁻¹ (FWHM). The spectra were extracted using the automated program, MAKEE, written by Tom Barlow. Misawa et al. (2007a) used the spectra of 37 of these quasars, after removing spectra of three quasars that cover only the Lyα forest region, and identified 39 candidate intrinsic systems (28 reliable and 11 possible systems) by partial coverage analysis. However, most spectra used in Misawa et al. (2007a) were produced by combining multi-epoch spectra in order to increase the S/N. In this paper, we used 4 NAL quasars (HE0130−4021, Q1009+2956, HS1700+6416, and HS1946+7658 out of the 37 quasars above because their original spectra (i.e., before combining data from multiple epochs) were available to us. We also used a spectrum of the mini-BAL quasar (UM675) in one epoch. The spectrum was obtained by Fred Hamann, using Keck/HIRES with a 115 (Hamann et al. 1997b), who kindly provided it to us.

Spectra of seven of the quasars studied here were obtained from the VLT/UVES archive. Narayanan et al. (2007) retrieved all R ∼ 45,000 spectra made available before 2006 June, and made a catalog of Mg ii absorption systems in 81 quasar spectra. Since most of these quasars were originally observed for studies of intervening absorbers, they should not have a particular bias toward or against intrinsic NALs and mini-BALs. In this catalog, we found four mini-BAL systems in the spectra of HE0151−4326, Q1157+014, HE1341−1020, and Q2343+125, as well as intrinsic NAL systems in two quasars (HE0130−4021 and HE0940−1050). Some of the mini-BAL quasars above were already discovered in past work (Hamann et al. 1997a; Menou et al. 2001). In addition to these archival data, we re-observed three NAL quasars (HE0130−4021, HE0940−1050, and Q1009+2956) and four mini-BAL quasars (HE0151−4326, Q1157+014, HE1341−1020, and Q2343+125) in 2007, using VLT/UVES with appropriate standard setups for each target to cover the wavelength from Lyα to C iv λλ1548,1551 with only a few exceptions. We used a 10 slit (yielding R = 40, 000; 7.5 km s⁻¹) and 2 × 2 CCD binning. Neither the Atmospheric Dispersion Compensator (ADC) nor the image-slicer were used. We reduced the data following the procedure of Narayanan et al. (2007) using the European Southern Observatory provided MIDAS pipeline. We also performed helio-centric velocity correction and air-to-vacuum wavelength correction.

We obtained high-resolution spectra of one NAL quasar (HS1700+6416) that was identified in Misawa et al. (2007a) and two mini-BAL quasars (UM675 and HS1603+3820) from the literature (Hamann et al. 1997b; Dobrzycki et al. 1999) using Subaru/HDS. We chose these quasars because of their good visibility from the Subaru telescope and their brightness, m_V ⩽ 17. We used a 10 slit (yielding R ∼ 36,000, 8.33 km s⁻¹), the ADC, and the red-sensitive grating. The CCD binning was set to 2 × 1. We used non-standard setups to cover Lyα, N v λλ1239,1243, Si iv λλ1394,1403, and C iv λλ1548,1551. We reduced the data in a standard manner with the IRAF software.⁶ Wavelength calibration was performed using the spectrum of a Th–Ar lamp. Because the blaze profile function of each echelle order changes with time, we could not perform flux calibration using the spectrum of a standard star. Therefore we directly fitted the continuum, which also includes substantial contributions from broad emission lines, with a third-order cubic spline function. Around heavily absorbed regions, in which direct continuum fitting was difficult, we used the interpolation technique introduced in Misawa et al. (2003). We have already confirmed the validity of this technique by applying it to a stellar spectrum.

To enhance the S/N of the spectra, all available observations of a particular target, taken within 2 weeks (14 days) of each other, were combined into a single spectrum, representing a single epoch. Although some BAL features show variability on very short timescales (Δt_rest ⩽ 0.1 yr), the variation probability is much smaller than that for longer time intervals (e.g., Capellupo et al. 2013). Therefore we prioritize increasing data quality above investigating short-term variability (i.e., combining as many spectra as possible to increase the S/N). In total, we have monitored 12 intrinsic NALs in 5 quasars and 7 mini-BAL in 6 quasars for 4–12 yr (∼1–3.5 yr in the quasar rest-frame). Each was observed with one or more of the telescopes/spectrographs discussed above. The observations are summarized in Table 2.

For the data analyses in the following sections, we normalized all spectra as follows. We fit the VLT/UVES spectra (including both the continuum and the broad emission lines) with cubic spline functions and divided the original spectra by these functions. While fitting, we avoid absorbed regions so that the fit interpolates across them. An example of continuum fits around the C iv mini-BAL of the VLT/UVES spectrum of Q2343+125 (Epoch 1) is shown in Figure 1. Because the continuum level usually depends on the regions used for the fit, we experimented with three different fit regions. The uncertainty from the continuum placement will be discussed in Section 3.1. For Keck/HIRES and Subaru/HDS, we separately fit each echelle order of the spectra with a cubic spline function that includes the instrumental blaze function as well as the continuum and the broad emission lines. This was necessary because flux calibration was not applied to these spectra. (i.e., each echelle order is not smoothly connected to adjacent orders.)

Zoom In Zoom Out Reset image size

The normalized profiles of Lyα, N v λλ1239,1243, Si iv λλ1394,1403, and C iv λλ1548,1551, superimposed for all epochs of observation (listed in Table 2), of the 12 intrinsic NALs and 7 mini-BALs are displayed in Figures 2–13, if they are covered by the observed spectra. The different epochs are represented by curves of different colors, with the first to sixth epochs shown as black, red, green, blue, light-blue, and purple, respectively. The velocity scale is relative to the flux-weighted center of a line, calculated by Equation (4) of Misawa et al. (2007a). For this display the mini-BAL spectra, only, were resampled every 0.3 Å because their S/N are relatively low and their profiles are broad enough that no information is lost. The Lyα and other metal absorption lines of mini-BALs are omitted from this plot if they are badly blended with the Lyα forest and/or if they are observed with Subaru/HDS, because in that case the continuum fit is too uncertain to allow a variability analysis.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Our initial inspection of the normalized spectra shown in Figures 2–13 did not reveal any substantial variations in the profiles of either the NALs or the mini-BALs, i.e., no changes in the profile asymmetries or the velocities of the troughs were discernible. However, changes in the strengths of the mini-BALs were quite evident. To quantify changes in line strengths, we measured the EWs of the blue components of NALs and the entire doublet regions of mini-BALs in each of the observed epochs for each transition, i.e., C iv, Si iv, and N v. We list these measurements in Table 3 along with their uncertainties. For the mini-BALs, the doublet members often suffer self-blending and cannot be reliably separated. We calculated the 1σ error bar on the measured EW, σ_EW, as the sum of contributions from uncertainties in the intensities of individual pixels in the spectrum and from uncertainties in the placement of the continuum: $\sigma _{{{\rm EW}}}=(\sigma _{\rm pix}^2+\sigma _{\rm cont}^2)^{1/2}$. The first term was evaluated by summing in quadrature the EW error bars from the N individual pixels making up the absorption trough (see horizontal arrows in Figures 2–13)

$$\begin{equation} \sigma _{\rm pix}^2= \sum _{i=1}^N \left(\sigma _i \, \Delta \lambda _i \right)^2, \end{equation} \tag{ 1 }$$

where σ_i is the error in the normalized intensity at pixel i and Δλ_i is the width of that pixel (in Å). To compute the uncertainty resulting from the continuum placement, we assumed that it is proportional to the product of the width of the line near the continuum level, λ_max − λ_min (see horizontal arrows in Figures 2–13), and the characteristic uncertainty of the continuum next to the line, σ_f. In other words, we supposed that the fitted continuum level in the spectrum could fluctuate by an amount of the order of ±σ_f, sweeping out an area of the order of ±(λ_max − λ_min)σ_f, which sets σ_cont. The fluctuations in the normalized continuum level are σ_f/f, where f is the observed flux density, which is the inverse of the S/N. Hence, we may write

$$\begin{equation} \sigma _{\rm cont} = A\, (\lambda _{\rm max}-\lambda _{\rm min})\,{\sigma _f\over f} = {A\, (\lambda _{\rm max}-\lambda _{\rm min})\over {\rm S/N}}. \end{equation} \tag{ 2 }$$

In the above expression, we have included a "calibration parameter," A, whose value we found empirically as follows. We carried out several experiments in which we fitted the continuum around some NALs and mini-BALs multiple times, choosing different regions around them each time, so as to determine the amount by which the fitted level could fluctuate. We then compared the variation in EW measured after different continuum fits to that given by Equation (2) and found that A ≈ 0.5 very uniformly (see Figure 1). Thus we adopted Equation (2) with A = 0.5 as an estimate of the EW uncertainty resulting from continuum placement errors. We find that the continuum placement errors are the dominant source of uncertainty in the measured EWs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To assess the significance of observed variations in the EW we computed the change in EW between two epochs as ΔEW = EW₂ − EW₁ (where Epoch 1 is earlier than Epoch 2) and its uncertainty as $\sigma _{\Delta {{{\rm EW}}}} = ({\sigma ^2_{{{{\rm EW}}}1}+\sigma ^2_{{{{\rm EW}}}2}})^{1/2}$ and defined the variation significance as

$$\begin{equation} S_{\rm {{{\rm EW}}}} \equiv {\Delta {{{\rm EW}}}\over \sigma _{\Delta {{{\rm EW}}}}} ={{{{\rm EW}}}_2 - {{{\rm EW}}}_1 \over \left(\sigma ^2_{{{{\rm EW}}}1}+\sigma ^2_{{{{\rm EW}}}2}\right)^{1/2}}. \end{equation} \tag{ 3 }$$

Significant variability is seen preferentially in the mini-BAL systems (see Figures 14–19), but not in NALs. The greatest changes for NALs are seen from E1 → E2 of N v in the v_shift ∼ 452 km s⁻¹ system of Q1009+2956 and from E2 → E4 of N v in the v_shift ∼ 767 km s⁻¹ system of HS1700+6416. In both of these cases, the variation significance is between 1.5σ and 2σ, and this is consistent with expectations in a sample of this size that does not experience variability. It is also consistent with variations in the EWs of intervening NALs caused by measurement errors. We tabulate the measured variation significance for mini-BALs in Table 4 for the shortest time intervals where a variation was detected (in the case of Q2343+125 we only detect variability at a significance level of <2σ but we include this object in the table for completeness). Distributions of absolute variation significance, |S_EW|, as a function of EW and time interval are shown in Figures 20–23 for NALs and mini-BALs, respectively. Histograms of |S_EW|, measured for all unique pairs of observing epochs, are also shown in Figures 24 and 25.

Five of the six mini-BAL systems in our sample show significant variability at greater than |S_EW| > 2 in at least one transition (in the sixth quasar, Q2343+125, the C iv line varied at |S_EW| < 2; see Table 4). Out of 52 unique pairs of epochs plotted in the histogram of Figure 25, four are in the range 3 < |S_EW| < 5 while another 6 are in the range |S_EW| > 5. For NALs, we compare our sample with intervening NALs that are detected in the same spectra but are classified as intervening NALs with coverage fractions of unity (Misawa et al. 2007a). In Figure 24, we compare the distribution of |S_EW| of intrinsic NALs to that of intervening NALs. This comparison shows the two distributions to be very similar and reinforces our conclusion that none of the intrinsic NALs in our sample have varied significantly.

In Figure 26, we plot the change in EW of mini-BALs for every unique pair of observing epochs, |ΔEW|, as a function of the rest-frame time interval, Δt_rest. The EW changes are larger for longer time intervals for lines from all observed ions, C iv, N v, and Si iv. The trend persists if we normalize the EW changes by the average EW as shown in Figure 27. In other words, both the absolute and the fractional change in EW increase over longer time intervals (fractional variations of 50% or more occur on time intervals of a year or longer). Such a behavior has been noted for C iv BALs (Gibson et al. 2008; Capellupo et al. 2011, 2013; Filiz Ak et al. 2013). We find an analogous result, namely, when we detect significant variability in two transitions at the same time in the same system the EWs change in the same direction. We illustrate this behavior in Figure 28 where we plot the change in the Si iv or N v EW against that of C iv (see caption for error bars and other specific information). It is noteworthy, however, that we have never observed these three strong lines vary at the same time because one of them is either saturated or not detected. By comparing the data points in this figure with the straight line of unit slope included for reference, we conclude that the fractional variations of EW of lines in the same system are not necessarily proportional to each other.

If we assume that the observed changes in EW result from changes in the ionization state of the gas (see discussion in Section 1), we can set a lower limit to the electron density by taking the variability timescale, Δt_rest, to be an upper limit to the recombination time of the relevant ion, i.e., n_e ≳ (α Δt_rest)⁻¹, where n_e is the electron density and α is the recombination coefficient of the transition in question. We use the recombination coefficients for a gas temperature of 20,000 K (Arnaud & Rothenflug 1985; Hamann et al. 1995). It is not straightforward to infer a density from recombination to/from Si iv (Arnaud & Rothenflug 1985), therefore we perform the calculations only for C iv and N v. If the EWs have increased, we assume that C v → C iv and N vi → N v, while if the EWs have decreased, we assume that C iv → C iii and N v → N iv, based on the following argument. In all cases (with a single exception; see previous paragraph) of variable mini-BAL systems, absorption lines from all ions (i.e., Si iv, C iv, and N v, with ionization potentials of 45.1, 64.5, and 97.9 eV, respectively) get stronger and weaker together. In order for all three lines to increase and decrease together, the ionization parameter should be either log U ≲ −2.5 or log U ≳ −1.5 (Hamann 1997). The regime of low U can be rejected because we do not detect any absorption lines from low ionization species with ionization potentials between 14 and 33 eV corresponding to any mini-BAL systems (e.g., O i λ1302, Fe ii λ2344, Fe ii λ2383, Si ii λ1260, Si ii λ1527, Al ii λ1671, C ii λ1335, Al iii λ1855, Al iii λ1863, and Si iii). This suggests that log U ≳ −1.5 and that as the ionization parameter varies, the main ionization states of silicon, carbon, and nitrogen vary as follows: Si iv $\leftrightarrows$ Si v, C iv $\leftrightarrows$ C v, and N v $\leftrightarrows$ N vi. Under these conditions, we estimate lower limits on the electron density in the range 1.7 × 10³–4.0 × 10⁵ cm⁻³. We are only able to derive such limits for mini-BALs; since the NALs in our sample did not show significant variability, we are not able to apply this method to them. We list the results for individual objects in Table 4 and discuss them in more detail in Section 3.2

From the lower limits on the density of the mini-BAL gas, we then estimate upper limits to the distance of the absorber from the source of ionizing radiation by making use of the definition of the ionization parameter as the ratio of ionizing photon density to hydrogen number density at the illuminated face of the cloud:

$$\begin{equation} U\equiv {n_{\gamma }\over n_H} = {Q(H)\over 4\pi r^2\, c\, n_H}, \end{equation} \tag{ 4 }$$

where Q(H) is the emission rate of hydrogen-ionizing photons by the central engine and r is the distance of the absorber from that source. We assume that log U  >   − 1.5 so that C iv is the dominant ionization stage of carbon (Hamann et al. 1995, 1997b) and to satisfy the constraint that all the observed absorption lines increase and decrease together. We also assume that n_e ∼ n_H. Thus, we obtain Q(H) by scaling the bolometric luminosity, which we compute via L_bol = 4.4 λL_λ at λ = 1450 Å, based on the spectral energy distribution (SED) of Richards et al. (2006). This SED is a segmented power law of the form L_λ∝λ^α with α = −1.6 from 1000 Å to 10 μm, α = −0.4 from 10 to 1000 Å, and α = −1.1 from 0.1 to 10 Å, following Zheng et al. (1997) and Telfer et al. (2002). By integrating this SED, we find that Q(H) = 10^58.2 s⁻¹ for a quasar with a bolometric luminosity of 10⁴⁸ erg s⁻¹. Inserting the expression for Q(H) and limits on n_e and log U into Equation (4) and rearranging, we arrive at the following expression for the distance of the absorber from the continuum source:

$$\begin{equation} r \le 1.18\;L_{\rm 48}^{1/2}\;n_5^{-1/2}\; {\rm kpc}, \end{equation} \tag{ 5 }$$

where L₄₈ is the bolometric luminosity in units of 10⁴⁸ erg s⁻¹ and n₅ is the electron density in units of 10⁵ cm⁻³. Using Equation (5), we estimate upper limits on the distance of the absorber to be in the range 3–6 kpc (see Table 4 and Section 3.2). These values decrease by a factor of ∼4 if we use the SED of a typical radio-loud quasar (Mathews & Ferland 1987) or a high luminosity radio-quiet quasar (Dunn et al. 2010).

Zoom In Zoom Out Reset image size

Here we discuss the variable mini-BALs in more detail before moving on to a general discussion about their properties and the possible causes of variability.

• UM675 (z_em = 2.15), mini-BAL system at v_shift ∼ 1900 km s⁻¹ (z_abs ∼ 2.13).Time variability of the C iv and N v doublets over ⩽2.9 yr in the quasar rest frame was discovered in the mini-BAL absorber at z_abs ∼ 2.13 (i.e., v_shift ∼1900 km s⁻¹) in this radio-quiet quasar (Hamann et al. 1995). Additional observations indicated that the different ions can have different coverage fractions: C_f ∼ 0.45 for C iv, C_f ⩾ 0.83 for Lyα, and C_f ⩾ 0.6 for O vi (Hamann et al. 1997b).We collected two spectra taken with Keck/HIRES and Subaru/HDS in 1994 and 2005, respectively. As shown in Figure 8, N v and C iv got weaker at a >5σ significance over ∼3.5 yr in the quasar rest frame. Lyα also showed the same trend. The variation occurs across the entire profile, though perhaps fractionally less in the central narrower component, as seen in the spectrum of Q2343+125 (Hamann et al. 1997a). We cannot distinguish between a change in ionization and a change in coverage fraction as the cause of the variability.Summary. The C iv and N v profiles have weakened on timescales of several years in the quasar rest frame. This places constraints on density (n_e ⩾ 3.3 × 10³ cm⁻³), and thus distance (r ⩽ 4.0 kpc). The detection of Ne viii (Beaver et al. 1991) also suggests proximity to the quasar, although this line is not covered by our optical spectra.
• HE0151−4326 (z_em = 2.78), mini-BAL system at v_shift ∼ 11,300 and 8900 km s⁻¹ (z_abs ∼ 2.64 and 2.67).This system has not yet been studied in detail as an intrinsic absorption system. There are two distinct absorption systems at v_shift ∼ 8900 km s⁻¹ (corresponding to Δv ∼ 0 – 1000 km s⁻¹ in Figure 9) with C iv and Si iv detected, and at v_shift ∼ 11,300 km s⁻¹ (Δv ∼ −2800 to −1800 km s⁻¹) detected in C iv. In addition to these two systems, there is likely to be a weaker system at v_shift ∼ 6400 km s⁻¹ (Δv ∼ 1500–3000 km s⁻¹) with detected C iv.This mini-BAL system spans more than 5000 km s⁻¹ in velocity space, although it does not satisfy the criterion for being a BAL (which requires that the observed spectrum falls at least 10% below the model of continuum plus emission lines over a contiguous velocity interval of at least 2000 km s⁻¹).The first three epochs of observation are within one month, and neither C iv nor Si iv showed variation in EW at >5σ level, except for E2 → E3 of C iv for the v_shift ∼ 8900 km s⁻¹ system. The last epoch is separated from those first three by ∼2 yr in the quasar rest frame. The absorption is significantly stronger in the last epoch, again over the entire velocity range of observation. The Si iv is either not detected or very weak in the first three epochs, but then has appeared by the fourth epoch of observation in the component at Δv ∼ 800 km s⁻¹ (see Figure 9). This suggests that there has been an ionization change in this component at least, and not just a change in coverage fraction. To adequately test this hypothesis, we would need high quality spectra covering some higher ionization ions such as N v and O vi.Summary. The C iv profile for this absorption complex, which spans ∼5000 km s⁻¹, has increased in strength between observations separate by ∼2 yr in the quasar rest frame. Si iv has also appeared in one component. This is consistent with a lower degree of ionization at the later time, and places constraints on densities (n_e ⩾ 3.9 × 10³ cm⁻³), and thus distances (r ⩽ 6.3 kpc) for the v_shift ∼ 11,300 km s⁻¹ and 8900 km s⁻¹ systems.
• Q1157+014 (z_em = 2.00), mini-BAL system at v_shift ∼ 3000 km s⁻¹ (z_abs ∼ 1.97).Wright et al. (1979) first studied the spectrum of this quasar in detail, and noted that it has C iv and C iii] emission lines (although the Lyα emission line is absent or extremely weak) and several BALs such as Lyα, N v, and C iv. This system is classified as a BAL by the traditional definition because of the broad absorption profiles of C iv and N v. However, we regard it as a mini-BAL in this study because Si iv is not so strong as in typical BALs.We collected five epochs of observations with VLT/UVES from 2000 to 2007, covering C iv, Si iv, and N v at multiple epochs. The profiles are shown in Figure 10. The Si iv varies significantly on a timescale of ⩽223 days (⩽80 days in the quasar rest frame). The N v appears to get a bit weaker and the C iv to get a bit stronger over the interval, but the change is at less than a 5σ level except for E1 → E4 for C iv. The C iv and N v are clearly saturated over most of the profile, and so are not subject to variations in column density. A variation in coverage fraction would lead to a significant change in those EWs, since non-unity coverage fraction would cause these saturated profiles not to be black. Thus the lack of variability in the saturated C iv and N v suggests that an ionization change is responsible for the variability of the Si iv profile in this system.Summary. This system shows variability, but mostly in the weaker Si iv absorption, and not in the saturated C iv and N v. The most plausible explanation for this is that the coverage fraction remains similar, but that the ionization state changes, because a change in the C iv and N v column density won't make a significant difference in the profile if it is saturated (e.g., Capellupo et al. 2012). This would be consistent with these profiles having changed only in their unsaturated parts. Any strict constraints cannot be placed on density and distance because recombination from/to Si iv is not simple.
• HE1341−1020 (z_em = 2.134), mini-BAL system at v_shift ∼ 1300 km s⁻¹ (z_abs ∼ 2.12).We found this mini-BAL system with broad absorption features of C iv and N v in the catalog of Narayanan et al. (2007). As far as we know, this system has not been previously studied in detail as an intrinsic absorption system.We collected two epochs of observations with VLT/UVES in 2001 and 2007, which are shown in Figure 11. With the two available epochs we find that both lines show time variability on a timescale of ∼2 yr in the quasar rest frame. Both the C iv and the N v are stronger at the later epoch at more than 5σ significance level, and they are stronger over the entire profile. Si iv is not detected at either epoch. The profiles again have the appearance of narrower profiles superimposed on broad ones.Summary. The C iv and N v profiles have varied on a timescale of ∼2 yr in the quasar rest frame, with both ions becoming stronger over the full range in velocity. This is certainly consistent with a change in coverage fraction, but an ionization change also cannot be ruled out with only two ions available as constraints. If an ionization change is the cause of variability, we can place constraints on density (n_e ⩾ 3.0 × 10³ cm⁻³), and distance (r ⩽ 3.5 kpc). The absorption profiles show narrow components superimposed on broader ones.
• HS1603+3820 (z_em = 2.542), mini-BAL system at v_shift ∼ 9500 km s⁻¹ (z_abs ∼ 2.43).This system was first discovered by Dobrzycki et al. (1999), and then monitored with high-resolution spectra for ∼1.2 yr in the quasar rest frame (Misawa et al. 2005, 2007b). The C iv mini-BAL shows both partial coverage (Misawa et al. 2003) and time variability (Misawa et al. 2005, 2007b). The C iv profile changes across the entire range of velocity, ∼2000 km s⁻¹, at once (see Figure 12). Among three possible scenarios for the variation, (1) a gas motion, (2) a variable scattering material, and (3) a change of ionization condition, the first two cannot be applicable for this quasar (see Section 1). The preferred explanation of changing ionization parameter is likely to be caused by variable shielding material between the continuum source and the absorber, since a quasar of this luminosity does not vary enough to explain the change. Warm absorbers are a candidate for the shielding material as evidenced by X-ray observations of this quasar (Misawa et al. 2010; Rozanska et al. 2013).Summary. There is significant variability of C iv over the full velocity range on timescales as short as months in the quasar rest frame. The variability is likely to be caused by changing ionization parameter due to variable shielding material that modulates the quasar flux. We place constraints on the electron density (n_e ⩾ 1.6 × 10⁴ cm⁻³) and the absorber's distance from the flux source (r ⩽ 4.0 kpc).
• Q2343+125 (z_em = 2.515), mini-BAL system at v_shift ∼ 24, 400 km s⁻¹ (z_abs ∼ 2.24).Previously, Hamann et al. (1997a) detected C iv, N v, and Si iv mini-BALs at z_abs ∼ 2.24 (corresponding to v_shift ∼ 24,400 km s⁻¹) in this radio-quiet quasar spectrum. These lines showed both time variability (the lines weakened and then strengthened again by a factor of ∼4 on a timescale of ⩽0.3 yr in the quasar rest frame) and partial coverage (C_f ⩽ 0.2 for C iv).We collected additional spectra with VLT/UVES at three different epochs from 2002 to 2007. Because of the different wavelength coverage of the different observations, only C iv has multiple observations (Figure 13). Lyα and N v are blended with the Lyα forest. The EW measurements for C iv 1548 were 3.29 ± 0.69 Å, 2.04 ± 0.44 Å, and 2.69 ± 0.63 Å for observations 1, 2, and 3, respectively, indicating only negligible variability at less than the 2σ level. We do not detect any Si iv features, while Hamann et al. (1997a) found a weak Si iv mini-BAL. Several narrow components inside the broader absorption profiles of C iv (Hamann et al. 1997a) are not apparent in our lower-S/N spectra.Summary. In this v_shift ∼ 24,400 km s⁻¹ mini-BAL, C iv gets slightly weaker, and then stronger, with changes taking place in only 124 and 403 days in the quasar rest frame, but in < 2σ level. More detailed analysis is not possible due to there being multiple epoch observations only of C iv.

The observational results of our monitoring campaign, combined with earlier case studies (e.g., Misawa et al. 2007b; Hamann et al. 2011; Rodríguez Hidalgo et al. 2012 and references therein), suggest some similarities between the variability patterns observed in mini-BALs and BALs. In particular, we have found that the EWs of all mini-BALs in our sample vary on rest-frame timescales of months to years and have observed the following patterns (1) when different transitions in a mini-BAL system vary together, their EWs change in the same direction, (2) absolute and relative changes in EWs are larger over longer rest-frame time intervals, and (3) variations in the EWs of the lines are not accompanied by discernible changes in the profiles, i.e., different troughs in the profile vary together. In comparison (1) more than 50% of BALs vary over timescales of the order of a year in the quasar rest-frame with the probability of variation approaching unity on multi-year timescales (e.g., Capellupo et al. 2011, 2013), and (2) the EWs of the UV absorption lines from C iv, Si iv, and N v in BAL systems vary together and in the same direction, although not necessarily in proportion to each other (e.g., Filiz Ak et al. 2013). However, BAL profiles do vary as the EWs vary (e.g., Lundgren et al. 2007; Gibson et al. 2008; Capellupo et al. 2011; Filiz Ak et al. 2012, 2013) in contrast to what is observed for mini-BAL troughs. In view of the similarities, we consider the two variability mechanisms often discussed for BALs as explanations for the observed mini-BAL variability, gas motion across our sightline and change of ionization conditions (see Section 1 for references and more details)

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To evaluate the likelihood that the mini-BAL variations result from motion of the absorbing gas across the line of sight, we estimate the timescale for a parcel of gas to cross the cylinder of sight that encompasses either the continuum source or the BLR. We take the transverse speed of the absorber to be of the same order as the observed outflow speed (both are related to the free-fall speed, the only velocity scale in the problem), i.e., v ∼ 10⁴ km s⁻¹. We assume that the observed UV continuum is emitted from the inner parts of the accretion disk within a radius a_cont ∼ 10 r_g from the center, where r_g ≡ GM_•/c² is the gravitational radius and M_• is the mass of the black hole, i.e., a_cont ∼ 1.5 × 10¹⁵ M₉ cm, where M₉ = M_•/10⁹ M_☉. Using the bolometric luminosities from Table 4 and assuming that the accretion rates are close to the Eddington limit, we estimate the black hole masses to be in the range: M₉ = 2–15, which leads to continuum source radii in the range a_cont ∼ 3 × 10¹⁵–2.3 × 10¹⁶ cm and crossing times t_cross(a_cont) ∼ 0.09–0.7 yr. For the radius of the BLR, we adopt the relation between it and the continuum luminosity given by Kaspi et al. (2007), which leads to a_BLR ∼ 0.22–0.63 pc and crossing times of t_cross(a_BLR) ∼ 20–60 yr. The long crossing times of the BLR lead us to disfavor gas motion across the BLR cylinder of sight as the cause of variability. Moreover, a substantial fraction of the mini-BAL troughs in our sample are found at large outflow velocities, |v_shift| ≳ 5000 km s⁻¹, so that they do not overlap with the broad emission lines (i.e., the mini-BAL gas absorbs only continuum photons and the continuum source crossing time is the more relevant timescale). These considerations make this particular version of the scenario very unlikely. The crossing times of the continuum source for the more luminous quasars in our sample are on a par with the observed variability timescales and the amplitude of EW variations is largest for timescales just over a year, which is comparable to the time required to traverse the continuum source completely. However, for the less luminous quasars in our sample, we would have expected the EW variations to be faster. Moreover, all troughs in a mini-BAL system vary together when the EW varies, which is difficult to orchestrate unless the parcels of gas producing the mini-BAL troughs in a given system all cross the cylinder of sight to the continuum source at the same time. Thus, we disfavor this version of the scenario as the cause of variability, although we cannot decisively rule it out (see also the discussion in Rodríguez Hidalgo et al. 2011).

As argued by Misawa et al. (2007b), the fact that all troughs in a mini-BAL system vary together strongly suggests that the variability is the result of changes in the ionization state of the absorber, presumably because of a variable ionizing continuum. Filiz Ak et al. (2014) suggest that this mechanism may also operate in BALs. But since the continuum variability timescales of such luminous quasars are considerably longer than the observed variability of mini-BALs (e.g., Giveon et al. 1999; Hawkins 2001; Kaspi et al. 2007), one is led to invoke variable obscuration of the ionizing continuum seen by the absorbers by shielding material of variable column density. This screen could be the X-ray warm absorber (Gallagher et al. 2002, 2006), and indeed some X-ray absorbers toward mini-BAL quasars show this particular behavior (e.g., Giustini et al. 2011). In this scenario, an X-ray warm absorber lies along sight lines toward mini-BAL and BAL systems but not toward NAL absorbers. This conclusion is based on the measurement of high column densities from the X-ray spectra of quasars with BALs and mini-BALs but not from the spectra of quasars with intrinsic NALs (e.g., Chartas et al. 2009). Recent radiation-MHD simulations by Takeuchi et al. (2013) reproduce variable clumpy structures in warm absorbers with typical sizes or order 20 r_g and typical optical depth of the order of unity. They appear above heights of ∼500 r_g from the plane of the accretion disk but, interestingly, they are not found at very high latitudes, where sight lines towards intrinsic NAL absorbers have been suggested to be by Ganguly et al. (2001) and Misawa et al. (2008). The fact that these clumps have sizes comparable to the size of the continuum source, suggests that they can modulate the apparent size of the continuum source and, by extension, the coverage fraction of the absorbers. Thus, such clumpy structures provide a plausible explanation for the frequent variability of mini-BALs and the lack of frequent variability in intrinsic NALs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It is noteworthy that the intrinsic NALs we have monitored in this program did not show any variability. These were identified as intrinsic systems based on the fact that their C iv, N v, or Si iv doublets show the signature of partial coverage of the background source. Although the same signature is seen on some rare occasions in low ionization lines from intervening absorbers (presumably arising in cold and compact parcels of gas; see Kobayashi et al. 2002; Jones et al. 2010; Chen & Qin 2013), it has never been seen in high-ionization lines, such as C iv and N v, from intervening absorbers. Therefore, we are confident that the systems we have targeted are intrinsic. Moreover, variability has been observed in NALs in other monitoring campaigns, both at high and low redshifts, by Narayanan et al. (2004) and Wise et al. (2004). The NAL variability timescales and resulting lower limits on electron densities from those two studies are comparable to what we have found here for mini-BALs.

Narrow kinematic components are sometimes also found near the centers of mini-BALs. Examples in our sample can be seen at v_shift ∼ 24,400 km s⁻¹ in the C iv mini-BAL of HE1341−1020, at ∼1300 km s⁻¹ in the C iv and N v mini-BALs of UM675, and at ∼3000 km s⁻¹ in the Si iv mini-BAL of Q1157+014. Among these, we can monitor the narrow components in UM675 without complications from noise and line blending. The narrow C iv components of UM675 in Epochs 1 and 2 are shown in Figure 29, in which broader absorption troughs are fitted out by high-order spline functions. The total EWs (including both C iv λ1548 and C iv λ1551) of Epochs 1 and 2 are measured to be EW(E1) = 0.58 ± 0.02 and EW(E2) = 0.56 ± 0.01. Thus, they show no variability over a time interval of Δt_rest  ∼  3.5 yr. Hamann et al. (1997a) noted a similar behavior for the narrow core of the C iv mini-BAL in Q2343+125 and suggested the narrow components may not be physically associated to the quasar itself, but arising in the quasar host galaxy or foreground galaxy (although this quasar is one of our targets, we cannot monitor the variability of the narrow components because our spectra have lower S/N than those of Hamann et al. 1997a). However, narrow components appear to always coincide with the centers of mini-BAL profiles, which suggests a physical association between the absorbing media. One possible explanation for the behavior of the narrow cores of mini-BALs is that they originate in dense clumps of high optical depth, embedded in the mini-BAL flow. When the ionizing continuum varies, the broad mini-BAL troughs show an easily discernible response but the highly saturated, narrow cores do not.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The variability properties of mini-BALs and NALs reported here, in combination with previous reports in the literature (see references above), lead us to suggest the following possibilities for locations of the absorbing gas. These are inspired by numerical simulations of quasar outflows (see references in Section 1) and are a refinement of the picture suggested by Ganguly et al. (2001).

In the first scenario, both types of absorbers are located in an accretion-disk wind, as illustrated in Figure 30. The mini-BAL gas is in the form of filaments rising from the fast component of the outflow. They make up the interface between the dense BAL wind at low latitudes above the disk and a hotter, much more tenuous medium in the polar direction. As such, the mini-BAL filaments have a large outflow speed, comparable to that of the BAL gas and a substantial velocity gradient. Within those filaments there may exist small, dense clumps that are sometimes observed as narrow cores in mini-BAL troughs. Variability of the ionizing continuum incident on the mini-BAL filaments can change the optical depth of the mini-BAL troughs arising in the filaments but not those of their narrow cores which arise in the dense clumps with C iv optical depths of τ_C IV ≫ 1 (assuming their predicted hydrogen column densities of ≳ 10²⁰ cm⁻², log U ∼ −1.5, and solar abundances). In this scenario, the NALs and the narrow cores of mini-BALs arise in higher-density, smaller filaments in the same region of the wind or at higher latitudes. It is possible that the NAL filaments are denser than the mini-BAL filaments thus the NALs are less susceptible to fluctuations in the ionizing continuum. However, NALs could vary because of changes in the coverage fraction, which could come about as the apparent size of the background source changes because of fluctuations in the structure of the inner screen (see discussion at the end of Section 4.3).

Alternatively, the NAL gas may be located in the host galaxy, at large distances from the center, as we noted in Section 1. In this scenario, small NAL filaments may form from the interaction of the wind's blast wave with cold clouds in the host galaxy, as proposed by Faucher-Giguère et al. (2012). Radiative shocks resulting from this interaction can create thin, dense filaments whose properties are consistent with what is inferred from photoionization models for the NAL absorbers (e.g., Wu et al. 2010; Borguet et al. 2013) and are likely to have τ_C IV ≫ 1. Thus the variability of the NALs arising in these filaments is more likely to result from fluctuations in the coverage fraction than fluctuations in the ionizing continuum. This scenario is more likely to apply to associated NALs, with v_shift ⩽ 5000 km s⁻¹, as this range of blueshifts better matches the prediction of numerical models for the interaction of quasar outflows with the ISM of the host galaxy (Kurosawa & Proga 2009). NALs found at v_shift ≫ 5000 km s⁻¹ (e.g., 20,000–65,000 km s⁻¹; see Misawa et al. 2007a) are better explained by the previous scenario.

The underlying hypotheses in the above scenarios are that (1) NAL variations are more rare than those of mini-BALs (although when NALs do vary they do so on the same timescales as mini-BALs), and (2) that NAL variability is caused primarily by fluctuations in the coverage fraction while the variability of mini-BALs can be caused by fluctuations in both the coverage fraction and the ionizing continuum. Both of these hypotheses warrant further testing, which can be done by a more detailed analysis of the data we have presented in this paper as well as data collected from denser monitoring campaigns on select targets.