Tracking Outflow using Line-Locking (TOLL). II. Large Line-Locking Web identified in Quasar J151352+085555

The TOLL project’s parent sample of quasars is based on the catalog compiled by Chen et al. (2021b), which includes observations from VLT-UVES (Murphy et al., 2019) and Keck-HIRES (O’Meara et al., 2015). These spectra have resolutions ranging from $22,000\lesssim R\lesssim 71,000$ for VLT-UVES. In the Chen et al. (2021b) study, each candidate mini-broad absorption line (mini-BAL) system was carefully examined to ensure it was not contaminated by unrelated intervening absorption features, such as damped Ly$\alpha$ or Lyman limit systems originating from foreground galaxies or their extended halos (e.g., Prochaska et al., 2015; Berg et al., 2015; Prochaska et al., 2008). Our main goal is to study these line-locked features in order to gain deeper insights into the dynamics of quasar outflows. Thus, we conducted a thorough visual inspection of both the normalized and unnormalized spectra of all quasars in the catalog, and selected J151352+085555 for its promising potential to reveal line-locked systems.

Quasar J151352+085555 (also named Q1511+091, SDSS J151352.52+085555.7) is optically bright ($m_{\rm B}=18.0$) and has a emission redshift of $z=2.901$ measured from SDSS (Abolfathi et al., 2018; Pâris et al., 2018). It is classified as a C iv broad absorption line (BAL) quasar, and has a bolometric luminosity of $10^{14}L_{\odot}$ and virial black hole mass of $2.5\times 10^{10}M_{\odot}$ (Shen et al., 2011; Weedman et al., 2012). Sargent et al. (1988) studied its C iv absorption using the Palomar 5m Hale Telescope, and found a strong complex absorption line system at a absorption redshift of $z_{obs}\sim 2.85$. It was found to be radio intermediate from the VLA FIRST survey, having a radio loudness of 1.34, and was given the name of FIRST J151352.5+085554 (Helfand et al., 2015; Shaban et al., 2022a). A point source, CXOG J151352.5+085555, was detected in the X-ray band by Chandra near the optical position of quasar J151352+085555, with an variability amplitude about 50% and an average X-ray luminosity of $\rm{4\times 10^{45}~{}erg~{}s^{-1}}$ (0.5-7 kev, Wang et al., 2016; Shaban et al., 2022b; Quintin et al., 2024).

UVES is a dual-arm, grating cross-dispersed high-resolution optical echelle spectrograph mounted on the Nasmyth B focus of Unit Telescope 2 of the VLT (Dekker et al., 2000). The UVES observations of quasar J151352+085555 (PI: Srianand; ProgramID: 65.P-0038, 69.B-0108, 71.B-0136) were conducted in 2000, 2002 and 2003 under a medium seeing of 0.56 arcsec, with a total on-target exposure time of 52713 seconds (see also Murphy et al., 2019). Seventeen exposures were taken in the UVES standard configuration using the 0.8” and 1” slit, providing a spectral resolution of R$\sim$50,000 and wavelength coverage of 3284$-$10429 Å. Murphy et al. (2019) reduced the data using a modified version of the UVES reduction pipeline based on the optimally extracted method (Horne, 1986; Piskunov & Valenti, 2002; Zechmeister et al., 2014; Ma & Ge, 2019). The wavelength calibration is done using a single ThAr exposure (Murphy et al., 2007). We show the target spectrum at the rest-frame wavelengths in Figure 1. We did not detect significant time-variable absorption lines from the spectra taken about three years apart. Different observations then were combined to form a single spectrum with a dispersion of 2.5 km s^-1per pixel. The final continuum-normalized spectrum has a count-to-noise ratio of 19, 40, 77, 78, and 52 per pixel (2.5 km s^-1) near 3500, 4500, 5500, 6500, and 7500 Å in the observed frame.

We use a standard fitting procedure to extract all relevant absorption line systems from the quasar spectra (Chen et al., 2018, 2019, 2021b). The fitting process begins with a thorough examination of the continuum placement provided in the UVES archive spectra for each line of interest. In cases where the continuum appears to be poorly constrained or exhibits sharp variations compared to the unnormalized spectrum, we refit the continuum using a locally-constrained simple power law. Then we start to identify individual velocity components in the associated absorption line (AAL) complexes, beginning with the C iv $\lambda$1548, 1551 doublet. This doublet is typically strong in our target spectra, occupies wavelength range outside of the Lyman forest, and has a high signal-to-noise ratio. We then search for a wide range of other plausible lines at the corresponding redshifts to the C iv doublets, such as Si iv $\lambda$1394, 1403, N v $\lambda$1239, 1243, O vi $\lambda$1032, 1038, and the Lyman series. Upon identifying new components at the redshifts associated with other absorption lines, we proceeded to verify the presence of corresponding C iv lines in return. This process of mutual validation continued until we had thoroughly cataloged all the lines involved. Our spectra include the AAL complexes of C iv, Si iv, O vi, and N v doublets. In some cases, other lines help confirm the positions of C iv components. For instance, component 5 in Figure 2 are identified through Si iv. Table 1 lists all identified lines in the quasar spectra.

Our line-fitting procedure follows the approach detailed in Chen et al. (2018, 2019). We fit each candidate AAL for C iv, Si iv, O vi, N v and the Lyman series using a Gaussian optical depth profile:

where $\tau_{v}$ is the optical depth at velocity $v$, $\tau_{0}$ is the central optical depth, $v_{0}$ is the line center velocity, and $b$ is the Doppler parameter. The two lines in each doublet are fit simultaneously, sharing the same velocity shift and $b$ value. This procedure can also apply to the multiple lines in Lyman series, as they also share the same velocity shift and $b$ value.

We account for partial coverage of the background light source in the fitting, assuming a spatially uniform brightness of the source and a homogeneous absorbing medium. The observed intensity at velocity $v$ is then expressed as:

where $I_{0}$ is the continuum intensity, $I_{v}$ is the measured intensity at velocity $v$, and $C_{0}$ is the covering fraction along our line of sight, with $0<C_{0}\leq 1$ (Ganguly et al., 1999; Hamann et al., 1997; Barlow et al., 1997). We assume a constant covering fraction across the AAL profiles, attributed to the narrowness of the lines and the minimal variation in their lower extremities. But the value of $C_{0}$ can differ between lines.

The column densities for C iv, Si iv  O vi, and N v are derived from the fitted optical depths using the relation:

where $N_{i}$ is the ion column density, $f_{\mathrm{oscillator}}$ is the oscillator strength, $\lambda_{0}$ is the rest wavelength, and $b$ is the Doppler parameter.

While most fits are straightforward, certain lines present challenges due to blending with other absorption systems or unrelated lines. When multiple systems are blended, we fit the doublets simultaneously and resolve individual AALs only if they can be clearly distinguished visually. For regions containing unrelated absorption at different redshifts, we mask these areas before conduct the fitting procedure. Most O vi lines are heavily contaminated within the Ly$\alpha$ to Ly$\epsilon$ forests. We estimate their lower limits for conservation. For the components that show multiple Lyman lines, we try to fit as many Lyman lines as possible simultaneously to get the best constraints on the column densities and covering fractions.

The fitted line parameters and corresponding uncertainties of the C iv, Si iv, O vi, and N v doublets from the VLT-UVES spectrum of quasar J151352+085555 are summarized in Table 1. The table presents the measured quantities for each line, including the velocity shift $v$, line identification, rest wavelength, observed wavelength, Doppler parameter $b$, logarithm of the column density, and covering fraction $C_{0}$. These measurements are provided separately for the C iv, Si iv  O vi, and N v doublets. The last column contains notes regarding any line blends. The data are organized by redshift, corresponding to the component numbers listed in the first column, as shown in Figures 2, 3, 4, 5 and 6. The uncertainties reported for most parameters represent the $1\sigma$ errors obtained from the fitting procedure, primarily reflecting pixel-to-pixel noise in the spectra. For cases of significant blending, we employ the methods outlined in Section 3.2 of Chen et al. (2018) and Section 3.1 of Chen et al. (2019) to directly estimate parameter values or place limits on them.

Figure 2 displays our line profile fits for all C iv doublets included in the final catalog. The velocity shift $v$, covering fraction $C_{0}$, and column density $\log N$ are derived from Equations 1, 2 and 3 through fitting the C iv doublet line profiles. Figure 3, Figure 4, Figure 5 and Figure 6 illustrate the fitted profiles for the Si iv, N v, O vi, and Lyman AALs, respectively, as included in the final catalog.

We searched for line-locked pairs of absorption lines corresponding to the C iv, Si iv, O vi, and N v doublets in the quasar J151352+085555 spectrum, utilizing the component matrix in velocity space displayed in Figure 7. Line-locked pairs can be identified within the matrix by drawing a straight line with a slope of $+1$, with the intercept corresponding to the doublet velocity separation, which is approximately $498~{}\rm{km~{}s^{-1}}$ for C iv, $964~{}\rm{km~{}s^{-1}}$ for N v, $1939~{}\rm{km~{}s^{-1}}$ for Si iv, and $1650~{}\rm{km~{}s^{-1}}$ for O vi.

We define two absorbers as line-locked if their velocity difference meets the criterion:

where $v_{\rm sep}$ represents the velocity separation of the doublet, and $\rm{b_{1}}$ and $\rm{b_{2}}$ are the Doppler parameters of the two lines, respectively. Confirmed line-locked pairs are indicated using colored brackets in Figures 2, 3, 4 and 5.

In Figure 2, two line-locked pairs of C iv absorbers are identified: pairs (5, 7) and (8, 11), with separations matching the C iv $\lambda 1548,1551$ doublet. We also detect line-locking among other ionic doublets, including Si iv $\lambda 1393,1403$, N v $\lambda 1239,1243$, and O vi $\lambda 1032,1038$. For Si iv, two line-locked groups are found: (3, 6, 10) and (5, 9), as marked in red in Figure 3. In the case of N v, three successive doublets (3, 4, 6) form a line-locked complex, highlighted in Figure 4. We also identify three line-locked pairs in the O vi doublet: (1, 3, 5), (6, 9), and (7, 10).

The line-locking web can manifests as velocity-stabilized absorption features in quasar spectra, caused by the combined effects of multiple line-locking processes involving ions like C iv, N v, Si iv, and O vi, etc. Clouds in the web show distinct velocity separations matching the doublet spacings of common UV resonance transitions. From the component matrix in Figure 7, we have identified one of the largest line-locking web known to date. This web contains ten absorbers, which are interconnected through doublets of C iv, Si iv, N v, and O vi. Its structure is illustrated in the schematic shown in Figure 8. Such web configuration could occur by chance, we estimated the probability of alignment by chance following the approach of Srianand et al. (2002). We randomly populated 12 absorbers within the velocity space of -7000 to 0 km/s. The probability of detecting the arrangement depicted in Figure 8 is found to be $<10^{-8}$. This low chance alignment probability, together with the observation of multiple line-locked doublets in the same quasar (e.g., Ganguly & Brotherton, 2008; Hamann et al., 2011), supports line-locking as a natural consequence of radiative acceleration (Srianand et al., 2002; Juráňová et al., 2024). Traditionally, theoretical studies of line-locking have often employed simplified two-cloud models to explore the phenomenon (Lewis & Chelouche, 2023). The identification of the largest line-locking web calls for more sophisticated modeling. Potential interactions among the line-locked clouds in the web presents an interesting research topic for the future.

In this section, we explore the physical conditions of the absorber systems based on photoionization modeling. To gain insights into the ionization states, total column density, and element abundances of the absorbers, we employed the cloudy 2023 release (Chatzikos et al., 2023). Although a specific ionic column density can arise under various physical conditions, combining constraints from multiple ionic species helps narrow down the range of possible gas properties in the absorption systems. All 12 narrow absorption line (NAL) components identified (components 1 to 12) in Section 3.2 have measured column densities for at least two ion species, as listed in Table 1.

We run cloudy using a generic fixed H i column density of $\log N(\mbox{H\,{\sc i}})$(cm^-2) $=15$ that ensures that the clouds are optically thin in the Lyman continuum. This is justified by the measurements and upper limits on $N(\mbox{H\,{\sc i}})$ that we derive from the data (Table 1). In this optically thin regime, the column density ratios needed for our analysis do not depend on $N(\mbox{H\,{\sc i}})$ nor $N$(H). The calculations assume solar abundances and a standard quasar ionizing spectrum.

The ionization state of the absorbers is described by the ionization parameter, $U$, which is the dimensionless ratio of the number density of hydrogen-ionizing photons at the illuminated face of the clouds to the number density of hydrogen atoms. Figure 9 shows the calculated ionization fractions in important ions compared to well-measured column density ratios in the data (e.g., Si iv/C iv, C iv/N v, N v/O vi, etc.). These comparisons yield estimates of the ionization parameter $U$ (shown by the orange vertical lines in the figure) with uncertainties (orange shadows) based on the column density uncertainties listed in the data tables (see Section 4.3 in Chen et al. (2018) and Section 4.1.1 in Chen et al. (2019) for more details). In particular, if a component exhibiting absorption lines for multiple ions (e.g. Si iv, C iv, N v); in that case, the ion ratios (e.g. Si iv/C iv, C iv/N v, N v/Si iv) can yield distinct $U$ values. We derive a weighted mean (based on column densities uncertainties) to be the best $U$ value and a weighted error to be the measurement error.

Subsequently, we derive the total hydrogen column density, $N(\rm{H})$, for components 2, 4, 6 and 10 that have well-measured $N$(H i) values, and estimate upper or lower limits on $N$(H) for other components with only upper or lower limits on $N$(H i). This is accomplished by applying an ionization correction, H i/H, appropriate for the derived $U$ values to the $N$(H i) values measured from the data. The results are shown in Figure 10. The derived values span ranges of $\log N$(H) (cm^-2) from $\sim 18$ to $\sim 20.5$ and $\log U$ from $\sim-1$ to $\sim 0$, consistent with the findings of Srianand et al. (2002). In that study, the physical conditions of a single component (component e, corresponding to our component 5) were explored, yielding $-2.0\leq\log U\leq-0.6$ and $10^{19}$ cm^-2$\leq N(\textrm{H})\leq 10^{20}$ cm^-2, in agreement with our results of $\log U\sim-0.9$ and $N(\textrm{H})\lesssim 10^{20.3}$ cm^-2.

We then estimate the component metallicities using the general relation (see more details in Chen et al., 2018, 2019):

where $(\mbox{H/M})_{\odot}$ is the solar abundance ratio of hydrogen to some metal M, $N(\mbox{H\,{\sc i}})$ and $f(\mbox{H\,{\sc i}})$ are the column density and ionization fraction of H i, respectively, and $N(\mbox{M}_{i})$ and $f(\mbox{M}_{i})$ are the column density and ionization fraction of ion $\mbox{M}_{i}$ of metal $\rm{M}$, respectively. In particular, $N(\mbox{M}_{i})$ and $N(\mbox{H\,{\sc i}})$ are derived from our fitting process. We use our estimates of $U$ to determine the ionization correction factor $f(\mbox{H\,{\sc i}})$/$f(\rm{M}_{i})$ from the cloudy calculations, and then plug in measured values of the column densities to obtain the metal abundance [M/H].

The results for [Si/H], [C/H], [N/H] and/or [O/H], depending on lines available, are shown in Figure 11. The error bars shown in this figure are derive only from the measurement uncertainties in the corresponding ion column densities. Figure 11 shows a mixture of metallicity results for the five components. We have metallicity estimates for components 2, 4, 5, 6, and 10 because their H i column densities have been accurately measured or an upper limit has been estimated. For the other components, we have not displayed the limits because the lower limits on $N$(H i) result in high upper bounds for metallicity, well above solar, which do not offer meaningful constraints for our analysis.

Component 2 shows substantially sub-solar metallicity, very high velocity, and a sharp line profile, which are very different from the other 10 locked components. This indicates component 2 may form in a physically distinct region compared to the other locked components, likely farther away from the background continuum source-perhaps in the halo. This difference could explain why it is not locked with the other components. In components 6, we also find evidence for enhanced nitrogen abundance, with [N/C] $>0$, as reported in the literature (Petitjean et al., 1994, 2008; Fechner & Richter, 2009). For example, Fechner & Richter (2009) find 27$\%$ of the associated systems show clearly enhanced nitrogen. In Paper I, we also obtained reasonable [N/C] value for component 9 in Quasar J221531-174408. Using data from this study (component 4, 5, 6, 10) and Paper I (component 9), we find 20$\%$ (1 out of 5) NALs show evidence of enhanced nitrogen. Due to the small sample size of 5, this result should be considered preliminary and warrants further investigation with larger datasets to confirm its validity.

Assuming a tenth solar metallicity, cloud sizes ranging from 0.1 to 1 pc, and hydrogen column densities of $\log N(\textrm{H})$(cm^-2) ranging from 18 to 20.5 as derived from cloudy, we estimate the masses of typical absorbers to lie between $1\times 10^{-4}$ and $3$ $M_{\odot}$. Following the method of Paper I, we also estimate the total kinetic energy of the outflow system identified through line-locking, and evaluate its impact on the host galaxy’s evolution from the energetic perspective. Assuming typical distances of $\sim$10–100 pc for the line-locking clouds, we obtain kinetic energies $K\lesssim 1.65-165\times 10^{53}$ ergs, and total kinetic luminosity of $\dot{K}\lesssim(0.2-2)\times 10^{43}$ ergs s^-1 for this line-locking web. The resulting ratios between the kinetic energy rate and bolometric luminosity, $\dot{K}/L_{\rm Bol}\lesssim(0.5-5)\times 10^{-5}$, is two orders of magnitude lower than the threshold of 0.005 to 0.05 required for effective feedback to the host galaxies (Scannapieco & Oh, 2004; Di Matteo et al., 2005; Prochaska & Hennawi, 2009; Hopkins & Elvis, 2010). The conclusion is that the kinetic luminosity of the NAL outflow system is quite low compared to the bolometric luminosity of the quasar, which is consistent with the findings for the NAL outflow system in Quasar J221531-174408, as reported in our Paper I.

Quasar J151352+085555 is one of only four known quasars exhibiting line-locking (LL) signatures in at least C iv, Si iv, and N v doublets simultaneously. The other three are Q1303+308 (Turnshek et al., 1984; Foltz et al., 1987; Braun & Milgrom, 1989), J221531-174408 (Chen et al., 2024), and SDSS J092345+512710 (Lin & Lu, 2020b). As illustrated in Figure 8, quasar J151352+085555 exhibits one of the most complex LL structures reported in the literature. Ten out of the twelve identified NAL absorbers are linked through radiative forces, as evidenced by their velocity separations corresponding to the C iv, Si iv, O vi, and N v doublets. As discussed in Section 2, Srianand et al. (2002) is the first to identify the LL phenomena in quasar J151352+085555. They have investigated its LL properties using the UVES observation obtained at year 2000. Our analysis here uses the combined data obtained from UVES observation at year 2000, 2002, and 2003, and reveals some differences compared to the findings of Srianand et al. (2002).

While Srianand et al. (2002) focused on LL signatures associated with the O vi, N v, and C iv doublet splittings, we also identify two additional LL pairs corresponding to Si iv. Both studies have detected two pairs of LLs in C iv; however, we identify one different pair compared to their results. The first C iv LL pair reported by Srianand et al. (2002) (the b-c pair) is considered marginal in our analysis due to poor matching between the line profiles of the two components. The detected N v LL pairs are consistent between both works. Our study find two double LL pairs and one triple LL pair in O vi, while Srianand et al. (2002) detected only one triple LL pair. The discrepancies arise mainly from differences in data reduction methods and LL identification technique. Specifically, we have identified seven additional LL pairs compared to Srianand et al. (2002), thanks to the improved data reduction provied by Murphy et al. (2019) and the use of a different LL signature identification algorithm.

Srianand et al. (2002) have studied physical parameters of component e (corresponding to component 5 in this study) using a photoionization model. They reported a covering fraction of 0.4 and column density $\log N(H\mathrm{I})$ around 16.3, while we obtained values of $\gtrsim 0.4$ and $\lesssim 15.8$, respectively. They have shown the ionization parameter of component 5 should be in the range $-2.0<\log U<-0.6$, while we gave a value around -0.9. For the total hydrogen column density of component 5, Srianand et al. (2002) gave $19\lesssim\log N(H)\lesssim 20$, and we obtained a value of $\log N(H)\lesssim 20.3$. For all the other components presented in their Table 1, the covering fractions generally agree with results presented in our Table 1. In Srianand et al. (2002) they argued the distance of the absorbers to the background continuum source is between 0.1 pc to a few kpc, while we simply adopted typical values of 10-100 pc when estimating the total gas mass in the outflow in Section 3.3. We also placed constraints on the metallicities of five components to be equal to or less than solar, whereas Srianand et al. (2002) did not report any metallicity measurements. The above comparisons focusing on component 5 suggest both studies yielded similar line fitting and ionization results. And we can conclude that the main differences between the two studies are the LL structure identified, possibly due to different LL identification technique.

We find ten out of the twelve NALs identified in the J151352+085555 spectra appear line-locked, which represents a notably high percentage. This suggests that line-locking may be prevalent in quasar outflow, like suggested by Bowler et al. (2014), and Chen et al. (2021a). Although the specific physical mechanism is still under debate (see the discussion in Bowler et al., 2014), it is widely agreed that line-locked systems are related to the physical process of radiative acceleration (Milne, 1926; Scargle, 1973; Braun & Milgrom, 1989; Lewis & Chelouche, 2023; Dannen et al., 2024), where successive shielding of the outflowing clouds locks the clouds one by one in outflow velocity. Using simulations, Lewis & Chelouche (2023) suggests the clouds’ properties need to be fine-tuned to within 1% for the clouds to be line-locked. The case presented here, together with our previously studied cases (Chen et al., 2018, 2019, 2024), can be used for a preliminary statistical test of this scenario. We have compared the ion column density, coverage fraction, and $\log U$ between the ‘shadowed’ absorbers and ‘shadowing’ absorbers in Figure 12. We find that most ion column densities and ionization parameters of the shadowing clouds lie within a factor of $\sim 3$ of those of the shadowed clouds. Additionally, the coverage fractions of the shadowing clouds are comparable to those of the shadowed clouds. This generally supports the study of Lewis & Chelouche (2023), where they argue the two absorbers need to have similar physical properties. However, we also find exceptions, suggesting that clouds with quite different physical conditions can also be locked together, and fine-tuning is not always required.

One possible solution to the discrepancies between observational evidence and the theoretical prediction of Lewis & Chelouche (2023) is to add more clouds ($>10$) and multi-ionic LL ($>3$) to the modeling process, instead of using a simple two clouds one doublet model. Further theoretical modeling and detailed simulations are needed to reveal the mechanisms behind these line-locking phenomena.