Survey of Extremely High-velocity Outflows in Sloan Digital Sky Survey Quasars

To search systematically for C iv absorption that appears below the quasar continuum, we started by normalizing the quasar spectra. We have taken a conservative normalization approach: we selected the lowest possible fit through the continuum; thus, our absorption measurements might be lower limits in those cases where we underfit the continuum if the location of the "true" continuum is uncertain. Figure 1 shows an example of our systematic normalization. The majority of quasars show greater flux at shorter wavelengths in the UV/optical region, and composite quasar spectra are well approximated by a power law in this region, even in the far-UV (e.g., Vanden Berk et al. 2001; Telfer et al. 2002; Tilton et al. 2016). We used a simple power law (similar to what is shown in Figure 6 of Vanden Berk et al. 2001) anchored to three points in the quasar spectra (see Figure 1). These points are defined as the central value of the wavelength and the median value of the flux in four distinct wavelength regions. Regions A (rest-frame 1701–1725 Å) and B (rest-frame 1677–1701 Å) are located in spectral regions where we determined emission and absorption are typically not present after visual inspection and comparison to emission-line tables. Regions C (rest-frame 1280–1284 Å) and D (rest-frame 1415–1430 Å) were used to define the slope of the power law. As both C and D may sometimes be affected by emission and/or absorption, we adopted the following algorithm: (1) we used region C, together with A and B, to define the power law; (2) we compared this power-law fit at the midpoint of region D with the median value of the flux in region D; (3) if the difference between the two was not zero within 3σ (three times the median error in region D), we used region D, together with A and B, to define the power law instead. We ignored bad pixels with zero inverse variance.

Zoom In Zoom Out Reset image size

Besides the possibility of absorption being present, the spectral region between the Lyα and Si iv+O iv] emission lines may be complicated by additional emission lines that are only sometimes present and often weaker than Si iv+O iv] (such as O i and C ii; see Figure 1).⁷ In this paper, we did not attempt to fit these emission lines for each individual spectrum, but all normalizations were visually inspected.

For those cases where the systematic fit was judged to be unsuccessful (e.g., too low), we attempted individual normalizations, substituting or including additional anchor points throughout the whole spectrum. When the continuum between the Lyα and Si iv+O iv] emission lines was complex, we kept the spectrum in our sample only if the individually normalized fit at 1415–1725 Å was successful at defining a plausible continuum when extrapolated into this complex region; five spectra in which this was not the case were rejected from our sample. Additionally, we removed 12 spectra that lacked flux information for more than half the wavelength region of interest. In total, we rejected 17 cases, reducing the parent sample to 6743 quasar spectra.

The next step consisted in identifying any broad and non-shallow absorption features in the spectral region corresponding to C iv absorption outflowing at velocities between 0.1c and 0.2c. We restricted our search to cases where the continuum was absorbed: because we do not fit a pseudo-continuum to the continuum+emission lines (such as in Rodríguez Hidalgo et al. 2013), we only detect absorption below the continuum level. To search for absorption, we smoothed the spectra with a three-pixel boxcar. Measurements of absorption parameters described in Section 4, though, were carried out over the unsmoothed spectra.

Why broad absorption? Quasar spectra can include both absorption in the vicinity of the quasar (typically known as "intrinsic") and absorption that lies along the line of sight between the quasar and us, but is located outside the quasar environment ("intervening" absorption). Intrinsic and intervening narrow absorption is not distinguishable at the SDSS resolution with only one observation (high-resolution observations and variability studies can be used to identify their nature, however; Narayanan et al. 2004; Misawa et al. 2007). Intervening absorbers in rich and massive clusters of galaxies (e.g., Girardi et al. 1998; Venemans et al. 2007) show that the velocity dispersion of intervening absorbers can reach up to ∼1000–1200 km s⁻¹. Therefore, the broader the absorption, the more likely it is intrinsically related to the supermassive black hole phenomenon. To avoid intervening absorption, we searched for broad absorption with widths larger than 1000 km s⁻¹.

Why not broad but shallow absorption? If quasar spectra include broad and shallow absorption, our process for continuum normalization would fit through the absorption as if it were the continuum and the absorption would be undetectable. In order to detect shallow absorption with long widths, alternative methods for continuum normalization must be used (i.e., fitting quasar spectra templates). Within the method we used, we are not able to detect this type of absorption.

The Balnicity index (BI; Weymann et al. 1991) is one of the standard ways to search systematically for and characterize absorption. The BI represents an equivalent-width measurement: it is larger when the absorption has larger width and/or depth, and it is calculated by the following integral:

$$\begin{eqnarray}&&\mathrm{BI}=-{\int }_{a}^{b}\left[1-\displaystyle \frac{f(v)}{0.9}\right]C\,{dv},\end{eqnarray} \tag{ 1 }$$

where f(v) is the normalized flux (see Weymann et al. 1991). Dividing f(v) by 0.9 avoids detections of shallow absorption; the square bracket is only positive if the normalized flux values are lower than 0.9, thus absorption only contributes toward the BI if it reaches depths below 90% of the continuum flux. The C variable only holds two values: zero and one. Initially, it is set to zero and only becomes one when the term in the square bracket is continuously positive over a velocity interval of our choice. To find only broad absorption, we set this velocity width to be 1000 km s⁻¹ (see above). The values of a and b are the velocity integration limits. Typically, those limits are set to search only at velocities lower than 25,000 km s⁻¹ to avoid contaminating the BI measurement derived from the C iv absorption with other absorption that would typically accompany the C iv absorption, such as Si iv. The BI definition has its limitations, and different values of C, a, and b have also been proposed and used in different studies (see for example, Hall et al. 2002; Trump et al. 2006; Gibson et al. 2009). In our study, we set a and b to be 30,000 and 60,000 km s⁻¹, respectively, to flag any potential EHVO absorption. (We later determine whether the absorption is due to C iv or other ions by visually inspecting each spectrum; see Section 3.3.) BI values can be slightly contaminated with intervening and unrelated absorption, but given the much larger width of C iv intrinsic absorption, this represents a small portion of the total BI or equivalent-width measurement.

We calculated the BI as described above for all normalized spectra. For all spectra in which the BI was larger than zero, we carried out a careful visual inspection to select a secure EHVO sample. Potential reasons for rejecting spectra from the secure sample are explained below.

First, the broad absorption might not be due to C iv. While broad absorption appearing between the Si iv+O iv] and C iv emission lines is typically attributed to the ionic transition of C iv, absorption appearing between the Lyα and Si iv+O iv] emission lines can be caused instead by (1) Si iv absorption, outflowing at lower velocity, which is the most frequent occurrence of an alternative identification, or (2) O i, Si ii, and C ii absorption, also outflowing at lower velocities. In order to discriminate between other possible identifications and include in our sample only secure cases of extremely high-velocity outflowing C iv, we visually inspected every spectrum where absorption was detected and follow the methodology described below. Given the large widths of the absorption features, notice that it is not possible to use the difference in doublet separation to identify ionic transitions.

In quasar spectra, absorption due to Si iv can be rejected easily in most cases because Si iv absorption always has corresponding C iv absorption outflowing at similar speeds; there are no confirmed cases in the literature of quasar outflows where Si iv absorption is present without corresponding C iv. This agrees with the solar Si/C abundance (Grevesse & Sauval 1998; Asplund et al. 2009; Lodders 2010) and the similar ionization potentials of Si iv and C iv. We used a similar approach to determine any other possible identifications, given that Si iv and C iv are the predominant ions in these outflows.

Figure 2 shows an example where flagged absorption was rejected by visual inspection because it is Si iv and not C iv. All cases with flagged absorption were checked for potential alternative identifications by locating the wavelengths in the spectrum where the corresponding C iv absorption outflowing at similar speeds would be. Whenever the identification of absorption in the region of interest was ambiguous and could be attributed to any other ion accompanied by C iv, we discarded it and did not include it in our sample. This, again, will result in our survey being a conservative lower limit of the number of EHVO C iv in quasar spectra.

Zoom In Zoom Out Reset image size

Another possible reason for rejection was the blending of narrow absorption that can result in apparent broad absorption. During visual inspection, we rejected any absorption profiles shapes that resembled blended narrow absorption. Intervening damped-Lyα absorption would thus be excluded from our sample.

While many of these rejections were unequivocal, we found approximately 60 cases that potentially could be EHVO but that did not meet our standards for the current sample included in this paper (see Section 4). Figure 3 shows examples of each type of these "candidates": these cases were ambiguous due to (1—top panel in Figure 3) a combination of a difficult continuum placement and borderline depth or width of the absorption, which in combination results in cases where a different normalization would not result in flagged absorption. In some cases (2—middle panel in Figure 3), the identification could be Si iv due to the presence of C iv in a large range of corresponding velocities, although the profiles do not resemble each other so we suspect the absorption is not Si iv. Finally (3—bottom panel in Figure 3), some cases appear to be the result of either the blending of real broad absorption with narrow absorption or severe blending of only narrow absorption lines. We have not included any of these cases in our analysis below, but they will be available in our database. Our plan is to monitor these cases for potential future inclusion in our sample.

Zoom In Zoom Out Reset image size

We found 40 cases of quasar spectra with EHVO C iv absorption. Table 1 includes some information on these EHVO quasars: QSO name, SDSS DR9Q data collection information (plate, MJD and filter), z_em provided by DR9Q (z_PCA), and whenever available, improved redshifts from Hewett & Wild (2010) ${z}_{\mathrm{HW}10}$. Finally, we include whether the quasar would be classified as a BALQSO (see below). Figure 4 shows several examples where the EHVO C iv absorption is shaded in red. In those cases with sufficiently large redshift, we could search for other ions in the same outflow (for example, N v and O vi, shaded in green and blue in Figure 4, respectively). We discuss the search for other transitions below in Section 4.2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (40 panels)

Tarball of figure set images

Table 1.  Information about the EHVO Quasar Sample

QSO	Plate-MJD-fiber	${z}_{\mathrm{em},\mathrm{DR}9{\rm{Q}}}$	${z}_{\mathrm{em},\mathrm{HW}10}$	BALQSO?
J000154.90−004956.4	4216-55477-0166	2.8302 ± 0.0005	...	no
J001306.14+000431.8	4217-55478-0020	2.1643 ± 0.0004	2.169 ± 0.002	no
J005922.65+000301.4	3735-55209-0532	4.1822 ± 0.0013	4.177 ± 0.008	no
J012700.69−004559.1	4229-55501-0337	4.0826 ± 0.0027	4.097 ± 0.002	no
J014548.55−000812.5	4231-55444-0035	2.7903	2.805 ± 0.002	yes
J071843.50+391720.3	3655-55240-0388	2.6298 ± 0.0006	...	no
J073505.97+280327.0	4456-55537-0704	2.3243	...	no
J074711.14+273903.3	4452-55536-0214	4.1143 ± 0.0011	4.128 ± 0.003	no
J075240.18+092523.0	4511-55602-0415	3.0223 ± 0.0007	...	no
J075852.68+133530.8	4506-55568-0824	3.3734 ± 0.0014	3.386 ± 0.007	no
J081337.14+155705.4	4498-55615-0502	2.4498 ± 0.0007	...	no
J083304.73+415331.3	3808-55513-0892	2.3583 ± 0.0008	2.329 ± 0.007	no
J085825.71+005006.7	3815-55537-0910	2.8684 ± 0.0007	2.863 ± 0.003	no
J092125.97−023411.9	3766-55213-0046	2.8047 ± 0.0005	...	no
J094023.55+404703.2	4571-55629-0730	2.5839 ± 0.0007	...	no
J094258.03+005359.5	3826-55563-0860	2.4749	...	no
J095005.90+362455.2	4573-55587-0698	2.2442 ± 0.0004	2.237 ± 0.002	no
J095254.10+021932.8	4743-55645-0118	2.1475	2.153 ± 0.002	no
J095603.47+382517.2	4570-55623-0296	2.8962 ± 0.0009	...	yes
J100400.20+372551.9	4567-55589-0454	2.5347	...	yes
J103456.31+035859.4	4772-55654-0106	3.3867 ± 0.0007	3.388 ± 0.002	no
J110841.93+031735.1	4741-55704-0786	3.3478 ± 0.0009	...	yes
J111154.35+372321.2	4622-55629-0678	2.0747 ± 0.0003	2.074 ± 0.002	no
J113000.22+344625.9	4619-55599-0854	3.6188 ± 0.0008	3.615 ± 0.004	no
J120609.69+004522.6	3844-55321-0920	2.6905 ± 0.0006	2.691 ± 0.003	no
J123056.28+345201.7	3968-55590-0464	3.4913 ± 0.0010	3.493 ± 0.003	no
J131307.93+065349.0	4840-55690-0576	2.9136	...	no
J133150.25+393417.7	4708-55704-0418	1.9996 ± 0.0005	1.991 ± 0.003	yes
J133752.36−011924.7	4046-55605-0085	2.7999	2.835 ± 0.005	yes
J135203.03+333938.3	3861-55274-0582	3.0742 ± 0.0014	3.079 ± 0.004	no
J144356.21+062539.2	4858-55686-0392	2.6063 ± 0.0006	...	yes
J150339.76+183423.9	3957-55664-0568	2.1848 ± 0.0003	2.184 ± 0.002	no
J151016.40+034200.0	4776-55652-0216	2.2468 ± 0.0006	...	no
J162445.03+271418.7	5006-55706-0846	4.4572 ± 0.0033	4.478 ± 0.003	no
J162747.14+192639.7	4060-55359-0940	2.4541 ± 0.0006	...	no
J164653.72+243942.2	4181-55685-0543	3.0329 ± 0.0006	3.040 ± 0.002	no
J165436.85+222733.8	4178-55653-0608	4.6910 ± 0.0032	4.708 ± 0.003	no
J171800.39+314203.9	4998-55722-0306	2.4677 ± 0.0005	...	no
J222559.52−004157.5	4202-55445-0258	2.7721 ± 0.0004	2.776 ± 0.002	no
J231227.48+005231.7	4209-55478-0712	3.5710 ± 0.0011	...	no

Note. Information on the 40 EHVO quasars found. QSO column includes the SDSS name, and we have also included the plate-MJD-fiber. The redshifts included are the z_PCA provided by SDSS (${z}_{\mathrm{em},\mathrm{DR}9{\rm{Q}}}$), with errors whenever available (see Pâris et al. 2012), and the improved redshift from Hewett & Wild (2010) whenever available (${z}_{\mathrm{em},\mathrm{HW}10}$). The BALQSO? column indicates whether the quasar is a C iv BALQSO in the traditional BI definition, defined at lower velocities.

Download table as:  ASCIITypeset image

Seven of the EHVO quasars are BALQSOs in the typical definition based on absorption present at lower velocities—between 5000 and 30,000 km s⁻¹. We have indicated this in Table 1 in the "BALQSO?" column. However, the BI measurement included in Table 2 is the one calculated at extremely high velocities only (see Section 3.2). The small number of EHVO quasars in our sample that are also traditional BALQSOs is not necessarily significant: in our goal of compiling a list of secure cases of EHVO quasars, we have rejected (or downgraded to potential candidate status) all of those cases where the potential EHVO absorption could be identified instead as low-velocity Si iv due to the presence of corresponding C iv absorption for a large fraction of the trough (see Figures 2 and 3 in Section 3.3). Some of these cases could have C iv present both at low and high velocities, however, and a large fraction of these would be BALQSOs. Thus, we might have systematically removed BALQSOs with potential C iv EHVO absorption.

Table 2 includes absorption information on all of the 40 EHVO quasar spectra. The BI_EHVO was calculated per quasar as described in Section 3.2. The upper and lower velocity limits for each EHVO, v_max and v_min, respectively, were determined at the wavelengths where the absorption crosses 90% of the normalized flux level; these are included for each EHVO absorption feature present in the spectrum. Six of the 40 quasars show two or more absorption troughs in the same spectrum outflowing at speeds larger than 30,000 km s⁻¹, making a total of 48 separated absorption features in the 40 quasar spectra. We consider those that contribute separately toward the BI value to be distinct absorption features, as explained in Section 3.2, and have different values of v_min and v_max. Equivalent widths (EW) are the integrated values of the absorption between the v_min and the v_max limits. Depth measurements were obtained as 1 minus the local minimum flux value of the trough avoiding noise spikes and unrelated narrow absorption; narrow lines contaminating the EHVO trough were masked prior to this calculation. All of our absorption measurements are carried out using the unsmoothed spectra. Errors on BI, v_max, v_min, EW, and depth are mostly influenced by the location of our continuum fit (a power law, see Section 3.1) as it will shift the location of the normalized flux level; we estimated these errors by raising and lowering the normalized continuum by an amount that would place the new continuum fit within the spectrum error (typically by 5% of the normalized flux) and considering the recalculated measurements as 3σ deviations. The resulting BI and EW σ errors are ∼20% of their values—typically near 100 km s⁻¹ for our BI median value of 400 km s⁻¹ and EW median value of 300 km s⁻¹.⁸ Typical v_min and v_max errors are ∼200 km s⁻¹. Values in Table 2 are rounded to reflect significant figures.

We caution against using either 40/6743 (secure cases) or (40+60)/6743 (secure+candidate cases) as estimates for the percentage of EHVO quasars among the 6743 parent sample quasar spectra. In this paper, we are presenting a sample of secure EHVO cases, and we have not included any potentially conflicting cases where the absorption is located on top of the emission or where the absorption might correspond to any other transition. Thus, any estimate derived from the frequency of the cases found in this sample is a very conservative lower limit of the percentage of EHVO cases in quasar spectra.

4.2.1. N v and O vi

4.2.2. Lyα

4.2.3. Si iv

Photoionization is the most important factor in determining quasar broad emission-line fluxes. Photoionization should affect the strength of broad absorption lines as well as broad emission lines. Although broad emission- and absorption-line strengths should be correlated at some level, those correlations will be smeared due to light travel time effects on emission lines and density-dependent response times of absorption lines.

The strength of the He ii λ1640.42 emission line is known to be related to the presence of soft X-ray continuum emission. Casebeer et al. (2006) discussed the dependence between different spectra energy distributions (SEDs) and the strength of the different emission lines including He ii. They showed that the He ii emission line is stronger for harder SEDs—in other words, when a larger fraction of higher-energy photons is emitted. In those conditions, observing strong winds is less likely because the gas may be too ionized to produce observable absorption. Indeed, Richards et al. (2011) showed that BAL-type quasars have typically weaker He ii emission in their composite spectra (their Figures 11 and 12) and predicted that quasars with strong He ii emission will be less likely to show outflows as absorption (see also Baskin et al. 2015).

We investigated the relative strength of the He ii λ1640.42 emission line in all our EHVO quasars. We follow a similar procedure to Baskin et al. (2013, 2015) to estimate the He ii EW by integrating the normalized flux in the λ_rest range between 1620 and 1650 Å. The Baskin et al. normalizing window (1700–1720 Å) overlaps with our normalization region A. Nonetheless, we divided each spectrum by the mean in that range again for an accurate comparison. We obtained a mean ${\mathrm{EW}}_{\mathrm{He}{\rm{II}}}$ of 0.505 Å. However, our quasars' spectra might have contamination at the red edge of the 1700–1720 Å range due to blueshifted emission lines. We decided to use both regions A and B (1677–1725 Å) as our normalizing window. With that normalization, we obtained a mean ${\mathrm{EW}}_{\mathrm{He}{\rm{II}}}$ of 0.83 Å, which is most similar to the lowest quartile value of 0.9 Å for low-z BAL quasars in Baskin et al. (2013) (see their Table 2) and to the value of 0.4 Å for the low-EW bin of high-z BAL quasars (see their Table 1). None of our values are larger than 5.4 Å, which was the median of their high-EW bin.

When measuring ${\mathrm{EW}}_{\mathrm{He}{\rm{II}}}$, we might be integrating Fe ii emission lines that tend to appear as a plateau redward of the C iv emission line. Weak He ii emission line would be indistinguishable from this Fe ii emission, but relatively strong He ii would show a peak at the corresponding wavelength, such as is observed in Figure 5 of Baskin et al. (2013), which would confirm the presence of He ii. We searched for this peak by measuring the mean flux in the 1630–1650 Å wavelength range as compared to the mean flux in the 1605–1625 Å range, combined with visual inspection of the spectra.

In Figure 5, we show examples of different strengths of potential He ii found in our analysis. In our sample, most cases show no indication of emission of He ii in that region (top panel), as both the EW and the relative flux measurements confirm. In ∼20% (7/40) of cases, we see some broad emission, confirmed by the He ii EW > 2 Å, but no peak at ∼1640 Å is observed, so it might be more likely to be Fe ii emission (middle panel). Only in three cases do we find an emission peak at the expected location of the He ii emission line (bottom panel in Figure 5), confirmed by either the EW or the relative mean flux. Notice that the O iii] λ1660.8,1666.2 emission lines are also present in this case.

Zoom In Zoom Out Reset image size

In conclusion, our results suggest that He ii is very weak in EHVO quasars, even weaker on average than in BAL quasars in general.

In this section, we study the parameter space of physical properties of the EHVO quasars in our sample versus their parent sample to learn whether they represent a special subgroup of quasars with particular properties. While DR9Q does not include measurements of bolometric luminosity (L_bol), black hole mass (M_BH), or Eddington ratio (${L}_{\mathrm{bol}}/{L}_{\mathrm{Edd}}$), those values have been published for quasars in SDSS Data Release 7 (DR7) in Shen et al. (2011). We cross-correlated our final sample with this catalog, which resulted on 21 EHVO quasars present in both DR7 and DR9. Table 4 includes the values directly from Shen et al. (2011) or derived from their data (see explanation below). The last line of Table 4 shows the median and dispersion values for each property.

Table 4.  DR7 Properties of DR9 EHVO Quasars

EHVO Quasar	$\mathrm{log}({L}_{\mathrm{bol}}/\mathrm{erg}\,{{\rm{s}}}^{-1})$	log (MBH/M⊙)	log (Lbol/${L}_{\mathrm{Edd}})$
001306.14+000431.8	47.065 ± 0.010	9.80 ± 0.06	−0.83
005922.65+000301.4	47.29 ± 0.03	9.43 ± 0.07	−0.35
012700.69−004559.1	47.584 ± 0.014	9.85 ± 0.05	−0.37
014548.55−000812.5	46.853 ± 0.015	9.88 ± 0.09	−1.21
074711.14+273903.3	47.750 ± 0.011	10.46 ± 0.15	−0.81
075852.68+133530.8	47.421 ± 0.010	10.02 ± 0.06	−0.70
083304.73+415331.3	46.740 ± 0.012	9.1 ± 0.5	−0.47
085825.71+005006.7	47.146 ± 0.012	9.30 ± 0.11	−0.17
095005.90+362455.2	46.958 ± 0.007	9.57 ± 0.08	−0.72
095254.10+021932.8	47.153 ± 0.006	9.56 ± 0.05	−0.51
103456.31+035859.4	47.625 ± 0.005	9.88 ± 0.05	−0.37
111154.35+372321.2	47.259 ± 0.005	10.10 ± 0.06	−0.94
113000.22+344625.9	47.316 ± 0.011	9.03 ± 0.08	0.24
123056.28+345201.7	47.470 ± 0.007	9.82 ± 0.04	−0.53
133150.25+393417.7	46.217 ± 0.019	9.76 ± 0.10	−1.65
133752.36−011924.7	46.68 ± 0.02	9.47 ± 0.13	−1.31
135203.03+333938.3	46.87 ± 0.03	9.9 ± 0.5	−1.15
162445.03+271418.7	47.540 ± 0.012	10.13 ± 0.17	−0.69
164653.72+243942.2	47.221 ± 0.016	10.02 ± 0.07	−0.90
165436.85+222733.8	47.745 ± 0.013	10.1 ± 0.6	−0.55
222559.52−004157.5	47.548 ± 0.007	9.50 ± 0.03	−0.02
Median,std	47.3,0.4	9.8,0.4	−0.7,0.4

Download table as:  ASCIITypeset image

We used the value and error of L_bol provided by Shen et al. (2011). For quasars with z_em ≥ 1.9, they calculated L_bol using the value of L₁₃₅₀ from their spectral fits and a bolometric correction BC₁₃₅₀ = 3.81 from the composite spectral energy distribution in Richards et al. (2006). Bolometric corrections were carried out for all DR7 quasars similarly, independently of whether or not the quasar was a BALQSO. No correction for intrinsic reddening was applied. BALQSOs are known to be redder than non-BALQSOs (see, e.g., Reichard et al. 2003; Gibson et al. 2009). Additionally, using L₁₃₅₀ to calculate L_bol is a potential issue as absorption might be present at that rest-frame wavelength for both BALQSOs and EHVO quasars. Thus, the true L_bol might be larger than the calculated L_bol from Shen et al. (2011). In fact, Shen et al. (2011) cautioned against the use of individual values of L_bol. However, we are using L_bol, as recommended, to study the average properties of these samples. We discuss in Section 4.4.1 how these issues might be affecting our findings, and in Section 4.5 we show the analysis of M_i[z = 2] that supports these results.

Our values and errors for M_BH in Table 4 are the adopted fiducial virial black hole mass in Shen et al. (2011). For quasars with z_em ≥ 1.9, Shen et al. (2011) used the C iv estimates from Vestergaard & Peterson (2006), which have been found to be somewhat overestimated at times, especially for those cases with large C iv blueshifts that might correspond to material that is not virialized (Shen et al. 2008). Quasars with large luminosities and BALQSOs are known to have particularly large C iv blueshifts (e.g., Pâris et al. 2012). Given that EHVO quasars overall show large values of L_bol relative to DR9Q (see Figure 6), EHVO quasars likely have overestimated M_BH values as well. While we do not know the C iv blueshift values for all of our EHVO quasars, we calculated a potential correction described in Coatman et al. (2017; their Equations (4) and (6)). We lack sufficient information on the C iv blueshifts to apply this correction to our sample of EHVOs, but for those cases for which we have C iv blueshifts (10 out of the 21 cases; from J. T. Allen & P. C. Hewett 2020, in preparation), we find that Coatman et al.'s (2017) correction for $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })$ would have been only −0.08, resulting in smaller M_BH values. We did not apply this correction to our measurements.

Zoom In Zoom Out Reset image size

Eddington ratios (log ${L}_{\mathrm{bol}}/{L}_{\mathrm{Edd}}$) were taken from Shen et al. (2011), who used the fiducial virial black hole mass to calculate them.

Throughout this section, we assume that the DR7 information is valid for EHVO quasars detected in DR9. We expect at most minor changes in the properties we study, as more luminous quasars tend to vary less than lower luminosity quasars (Kaspi et al. 2007). We defer the study of the variability of these quasars to a future paper.

4.4.1. Comparison to Parent Sample and BALQSOs

To compare the parameter space of properties described above for the 21 EHVO quasars with DR7 information and other classes of quasars, we cross-correlated the parent sample of quasars (6743) with the DR7 database of Shen et al. (2011) and found 2883 quasars in the parent sample with DR7 information.

In Section 2, we describe the cutoffs we performed over the original sample to obtain our parent sample. These cutoffs already select the brightest quasars within DR9Q due to the S/N constraint. Therefore, we are not comparing our sample to DR9Q as a whole. Notice as well that because we only include those DR9Q quasars in the parent sample with information in Shen et al. (2011), we are selecting a subsample of the parent sample that had previous observations in DR7.

To compare EHVO quasars to BALQSOs (quasars with strong and wide absorption at lower velocities), we selected cases within the parent sample that had been flagged as having BALs in Pâris et al. (2012). We constrained the BAL identification by selecting quasars that include all of the following requirements: (1) BAL flag from visual inspection, (2) absorption index AI > 0, and (3) $| {v}_{\min }| \gt 5,000$.

We used constraint (1) because the Pâris et al. (2012) BAL flags from visual inspection are robust at high S/N. While the authors caution against blind uses of the catalog at low S/N, this is not an issue in this case given that our parent sample has already a high-S/N cutoff. To compare similar widths of absorption in both samples, we used constraint (2). Our definition of BI is set for a width of 1000 km s⁻¹, similar to the AI definition in Pâris et al. (2012) and, originally, in Hall et al. (2002). Constraint (3) was used to remove possible contamination of "associated" absorption from our comparison between BALQSOs and EHVO quasars. Knigge et al. (2008) found a bimodal distribution of quasars because the AI definition, when applied from zero velocity, includes all absorption at redshifts close to the quasars' emission redshift. This so-called associated absorption might not be located near the central engine of the quasar but in the host galaxy or in neighboring galaxies, and their low-velocity absorption might represent a completely different type of phenomenon that is difficult to distinguish from low-velocity outflows. After applying all of these cutoffs, we found 444 BALQSOs in the DR9 parent sample; after cross-correlation with DR7, we found 185 BALQSOs. Three of the EHVO quasars are also BALQSOs.

Figures 6–8 show how the properties of the 21 DR7+DR9Q EHVO quasars compare to both the parent sample (2883 cases) and the BAL sample (185 cases) from DR7+DR9Q. Due to the smaller number of EHVO and BAL quasars, we have multiplied their frequencies in each bin by 10 and 2, respectively, so their distributions can be compared on the same scale.

EHVO quasars show larger values of L_bol and M_BH than their parent sample and BALQSOs. This can be observed in Figures 6 and 7, as well as in Table 5, where we show the median values of the three parameters for the parent sample, and the BALQSO and EHVO subsamples. The median value of L_bol in EHVO quasars exceeds both the median value in the parent sample and the median of the subsample of BALQSOs. The median of L_bol/L_Edd is similar for BALQSOs and EHVOs and the distributions overlap significantly (see Figure 8).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 5.  Median Values of the DR7 Properties for the Parent, BALQSO, and EHVO Quasar Samples

SDSS DR9 Quasar Name	Median log (Lbol/erg s−1)	Median log (MBH/M⊙)	Median log (Lbol/${L}_{\mathrm{Edd}})$
DR9Q parent sample	47.0	9.5	−0.5
BALQSOs	46.9	9.5	−0.7
EHVOs	47.3	9.8	−0.7

Download table as:  ASCIITypeset image

We performed statistical tests to determine the likelihood of obtaining these distributions of EHVO properties randomly from the parent and BALQSO samples. Specifically, we carried out Kolmogorov–Smirnov (K-S) tests between the parent and EHVO sample and between the BALQSO and EHVO distributions of L_bol, M_BH and L_Edd.

The results are given in Table 6. We find that EHVO quasars are statistically different from the parent sample and from BALQSOs in their distributions of L_bol and M_BH, but not in the distributions of Eddington ratios. It is especially relevant that we find such a small probability that the BALQSO and EHVO samples are derived from the same populations. While we find that our M_BH values might be slightly overestimated (see Section 4.4), our M_BH values for BALQSOs will be shifted in the same direction because they are also found to have large C iv blueshifts (e.g., Richards et al. 2011), making our relative comparison valid. We also note that the differences in M_BH between EHVOs and the parent or BALQSO samples are larger than can be explained purely by the partial dependence of M_BH (as calculated by Shen et al. 2011) on L_bol.

Table 6.  Results of the Statistical Analysis between EHVO Quasars, BALQSOs, and the Parent Sample for Values of L_bol, M_BH, and ${L}_{\mathrm{bol}}/{L}_{\mathrm{Edd}}$

		p-values
Sample	Lbol	MBH	${L}_{\mathrm{bol}}/{L}_{\mathrm{Edd}}$
EHVO—parent sample	2.7e−3	1.1e−3	0.18
EHVO—BALQSOs	8.8e−05	9.3e−03	0.70

Download table as:  ASCIITypeset image

The small differences found in L_bol are affected by how L_bol was determined, but little is known about the intrinsic properties of EHVO quasars, so we explore potential shifts of the L_bol distributions due to reddening and differential bolometric corrections. Reduction of L₁₃₅₀ by reddening of a quasar's spectrum by nuclear or host galaxy dust, or by broad absorption, would shift the L_bol calculated by Shen et al. (2011) to smaller values than the true L_bol value. Such a shift is likely seen in Figure 6 for the BALQSO sample, which is expected because BALQSOs are known to be redder than normal quasars (Reichard et al. 2003). If EHVO quasars have the same reddening distribution as BALQSOs, then EHVOs tend to have larger bolometric luminosities than BALQSOs. Alternatively, for reddening to explain the L_bol distribution for the EHVO sample, EHVO quasars would have to be less reddened than normal quasars. We have no ready explanation for why EHVO quasars would be less reddened than normal quasars, but we note that it might be natural for EHVOs to exhibit less reddening than BALQSOs if EHVOs are preferentially seen along high-latitude sight lines. Models that reproduce BALQSOs find outflowing gas along low-latitude sight lines that are more likely to intersect lines of sight through a dusty wind or torus (see, for example, Proga & Kallman 2004), while faster outflows are found at higher latitudes. Similarly, if absorption in EHVO quasars is reducing the value of L₁₃₅₀, it would be shifting the calculated L_bol for EHVO quasar to values smaller than the true L_bol value. Therefore, our finding that they show large values of L_bol instead is even more significant. In Section 4.5, we present an analysis of M_i[z = 2] that supports these results.

The possibility of differential bolometric corrections between EHVOs, BALQSOs, and non-BAL quasars cannot be ruled out from SDSS data alone. However, the fact that as a population BALQSOs have weaker He ii emission than non-BAL quasars, and that EHVOs have weaker He ii emission than BALQSOs (see Section 4.3), suggests that BALQSOs have weaker far-UV emission than non-BAL quasars, and that EHVOs have weaker far-UV emission than BALQSOs. Therefore, a differential bolometric correction relative to non-BAL quasars arising from different far-UV SEDs would be expected to have the same sign for BALQSOs and EHVOs. Smaller bolometric corrections could reconcile EHVO quasars with the parent population; however, a correction of the same sign will not reconcile both BALQSOs and EHVOs with the general quasar population (see Figure 6). Therefore, a differential bolometric correction is unlikely to be the sole explanation for the different L_bol distributions of both EHVOs and BALQSOS from non-BAL quasars. More work is needed to better determine the intrinsic properties of EHVO quasars.

BALQSOs and EHVO quasars appear to show distinct L_bol and M_BH properties. Because the major observational differences in their spectra are the minimum and maximum velocities of their C iv outflows, we investigated the relation between their outflow velocities and these properties. In Figure 9, we show the distribution of L_bol for EHVO and BALQSOs, where the BALQSOs have been divided into groups based on the minimum and maximum velocities of their C iv outflows. These velocities were extracted from Pâris et al. (2012).⁹ No progression toward higher L_bol is observed with increasing v_min in BALQSOs; all median values of $\mathrm{log}({L}_{\mathrm{bol}})$ are within 0.07 of 46.83, while the median for EHVO quasars is 47.26, and all p-values between subsets of values of $\mathrm{log}({L}_{\mathrm{bol}})$ for BALQSOs are larger than 0.5. A similar result is observed when selecting BALQSOs based on their v_max: all of their median values are within 0.02 of 46.88. We find the p-values between subsets selected based on v_max show smaller values overall than for v_min, but none of them are smaller than 0.03; we cannot conclude that those subpopulations of BALQSOs are distinct. On the other hand, all comparisons in $\mathrm{log}({L}_{\mathrm{bol}})$ between the EHVO population and subsets of BALQSOs larger than the EHVO sample have p-values smaller than 0.002.

Zoom In Zoom Out Reset image size

We do observe progressions in the values of M_BH with increasing v_min and v_max in BALQSOs. In Figure 10, we show the distributions of $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })$ for EHVO and BALQSOs, where the BALQSOs have been divided into groups based on the minimum and maximum velocities of their C iv outflows, similar to Figure 9. The distributions of $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })$ at lower velocities (both for v_min and v_max) are clearly distinct from the distribution of $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })$ values for quasars with larger velocities, and a progression toward larger values is observed in both cases. EHVO quasars fit correctly in this progression showing the largest values of black hole masses.

Zoom In Zoom Out Reset image size

To study this progression further, we subdivided the BALQSO population into smaller velocity bins. Due to small numbers, we did not divide into two subgroups the BALQSOs with 15,000 < v_min < 25,000 km s⁻¹ and those with 5000 < v_max < 15,000 km s⁻¹. Figure 11 shows the subsets of BALQSOs that are particularly distinct in each case. The largest gaps occur between BALQSOs with 5000 < v_min < 10,000 km s⁻¹ and those with larger v_min: the median of $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })$ for the latter population is 9.46, while all the subsets of BALQSOs with larger v_min show median values of $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })=9.59\pm 0.01$. Similarly, the largest gap at v_max occurs between BALQSOs with outflows at 20,000 < v_max < 25,000 km s⁻¹ and those with smaller v_max: the median of the latter is 9.63, while the other BALQSOs subsets show median values of $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })=9.44\pm 0.06$. Naturally, again, EHVO quasars fit as expected in this progression showing the largest median value of $\mathrm{log}({M}_{\mathrm{BH}}/{M}_{\odot })$ (9.82; see Tables 4 and 5).

Zoom In Zoom Out Reset image size

In summary, the distribution of M_BH of EHVO quasars is significantly distinct from the distribution of values for BALQSOs with the lowest velocities, but it is more similar to those BALQSOs with the largest velocities. The EHVO quasar population is distinct from BALQSOs with 5000 < v_min < 10,000 km s⁻¹ (p-value of 0.005), but when comparing it to BALQSOs with larger v_min, all p-values are larger than 0.07. Comparing the values for EHVO quasars to the subsets of BALQSOs in v_max, all p-values are smaller than 0.01 except for BALQSOs with outflows with and 20,000 < v_max < 25,000 km s⁻¹, for which the p-value is 0.13. Due to the small number of EHVO quasars in our current sample, we do not divide it into subsamples with different v_min or v_max to study potential trends within the velocity of the EHVO outflow, but this will be worth studying as we increase the number of EHVO quasars with future samples. Finally, note that EHVO quasars are distinct from all BALQSOs, independently of their v_min or v_max, in their distributions of L_bol values.

DR9Q includes information about an array of parameters and has the advantage that we can analyze the parameter distributions for the complete samples in our study: parent sample (6743 quasars), BALQSOs (444) and EHVO quasars (40). We obtained the values for several of these parameters (such as z_em, M_i[z = 2], and spectral indices) in the samples, and some of the results are shown in Figures 12 and 13.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The distribution in redshifts for EHVO quasars is distinct from that of BALQSOs and the parent sample. In Figure 12, we show the number of quasars per redshift bin scaled by the total number of objects in each category for the three populations. All redshifts were taken from DR9Q unless the Hewett & Wild (2010) redshifts were available. EHVO quasars are more predominant at larger redshifts. The median of the redshift value of EHVO quasars is z_em =  2.80, that of the parent sample is 2.42, and that of BALQSOs is 2.39. According to statistical tests, none of the populations are drawn from each other, as all the p-values are populations smaller than 9 × 10⁻⁴. Obviously, the cutoff at lower redshift (z_em > 1.9) is artificial; we imposed it to be able to find these type of outflows in the first place in the SDSS sample. However, beyond z_em > 1.9, the distribution of z_em for EHVO quasars does not follow the parent or the BALQSO distributions, both of which show a peak at z_em ∼ 2.3. Almost a third of the EHVO quasar population has z_em > 3.2, while less than 10% of the parent and BALQSO population show those values.

Figure 13 shows the fraction of BAL and EHVO quasars relative to the parent population in each redshift bin. EHVO quasars are less frequent at lower redshifts, but represent a larger fraction than BALQSOs at higher redshifts (z_em > 4), where there are only 52 SDSS DR9 quasars. Indeed, the quasar with the largest redshift in our parent sample from DR9Q shows an EHVO (1.0 value in Figure 13). This is not surprising, as at least one of the three DR9Q quasars with redshift z_em > 7 is also an EHVO quasar (see Section Section 5.2).

DR9Q also includes information on the absolute magnitude in the i band at z = 2, M_i[z = 2] (see more details in Pâris et al. 2012). M_i[z = 2] can be used as a proxy for bolometric luminosity, and it is available for all the quasars in our EHVO sample, and parent and BAL samples. The results are similar to the L_bol results shown in 4.4.1.

Figure 14 shows the distributions of M_i[z = 2] for the three studied samples, color-coded similarly to previous figures. For clarity, this figure does not include six outliers in the parent and BAL samples. As in Figure 6, the population of EHVO quasars shows more negative absolute magnitudes overall (median M_i[z = 2] = −28.07), and statistical tests show that it could not be randomly extracted from any of the other two populations (p-values < 4.5 × 10⁻⁵). Thus, EHVO quasars show smaller absolute (negative) magnitudes and larger luminosities overall, confirming what we found from the values of L_bol in the subsample that overlaps with DR7 (see Section 4.4.1).

Zoom In Zoom Out Reset image size

To contrast the potential effect of reddening in the spectra of the three quasar classifications, we use the power-law slope for the continuum fit around the C iv emission line provided by Shen et al. (2011), as it was the index available for all of the objects in our sample (${z}_{\mathrm{em}}\geqslant 1.9$). In terms of the distributions, BAL and EHVO quasars are distinct from non-BAL quasars, with EHVO quasars tending to lie somewhere between non-BAL and BAL quasars. A two-sample K-S test rejects the null hypotheses that BAL quasars have the same spectral index distribution as non-BAL quasars with p = 10⁻³² (formally) and that EHVO quasars have the same spectral index distribution as non-BAL quasars with p = 0.0018. The spectral index distributions of BAL and EHVO quasars are not different at a statistically significant level (p = 0.1037). Two caveats for the use of this index are (1) the wavelength region chosen is relatively narrow [1445, 1465] Å and [1700, 1705] Å, so it might not reflect the overall shape of the continuum, and (2) the Fe ii emission lines redward of the C iv emission were not subtracted in the spectrum prior to the continuum fitting. Therefore, we just use this spectral index as a comparison tool, but do not report numbers about individual EHVO quasars to avoid those numbers being used for the quasar's overall spectral characterization. Future work using composite spectra and including emission-line analysis will need to be carried out to define well the power-law spectral index of EHVO quasars.

DR9Q also includes information about the radio loudness of the quasars, as it has been cross-correlated with the Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) catalog (version 2008 July). Pâris et al. (2012) included the FIRST peak flux density at 20 cm (f_ν(20 cm), in mJy) for those quasars where FIRST had a detection within 20 of the quasar position.

We only find one EHVO quasar (1/40, corresponding to a fraction $2.5{ \% }_{-2.0 \% }^{+5.7 \% }$) where the value of f_ν(20 cm) was nonzero. The value is also quite small (f_ν(20 cm) = 1.08 mJy); for reference, more than 100 quasars in the parent sample have values larger than 50 mJy. For BALQSOs, $3.4{ \% }_{-0.9 \% }^{+1.1 \% }$ of quasars have values of f_ν(20 cm) larger than zero (15/444), but only one case has a value larger than 10 mJy.

Higher resolution may be required to determine the true radio nature of EHVO quasars, as radio-quiet quasars at low resolution have been found to show evidence for relativistic jets in such observations (Jarvis et al. 2019).

In a large parent sample (6743) of SDSS quasar spectra, with S/N > 10 and z_em ≥ 1.9, we have found 40 quasars with EHVOs (30,000 < v < 60,000 km s⁻¹), observed as C iv absorption in their spectra. We have investigated some of their properties as measured in the SDSS DR7 and DR9 and their relation to quasars with broad absorption at lower velocities (BALQSOs). Here we summarize the main results of this study.

• 1.  
  The 40 EHVO quasars show a large range of BI_EHVO (from 2 to 3600 km s⁻¹, using a minimum width of 1000 km s⁻¹ in the definition of BI_EHVO). The number of EHVO quasars includes only secure cases, and it should be considered a very conservative lower limit of EHVO cases in quasar spectra (see Sections 3 and 4).
• 2.  
  Seven of these EHVO quasars are also BALQSOs as flagged in DR9Q; this number must be considered a lower limit because we reject identifying as an EHVO any absorption that could be attributed to Si iv with corresponding C iv at lower velocity, and that excludes many potential BALQSOs from being identified as EHVOs (see Section 4).
• 3.  
  We find 48 EHVO absorption features in 40 quasar spectra. EHVO absorption shows a large range of v_min, v_max, and EW values, but depth values all lie between 0.14 and 0.71, with most values around 0.3 (see Section 4).
• 4.  
  Out of the 48 absorption features, 26 have confirmed or likely corresponding N v outflowing at similar speeds, and O vi, at least, is likely present in 6/48 cases as well. We do not observe any confirmed Lyα absorption in an EHVO (see Section 4.2).
• 5.  
  Most EHVOs show no signs of He ii emission in their spectra. In 20% of cases, we find some broad emission at those wavelengths, most likely due to Fe ii emission. Only three cases show narrow and strong emission that is likely due to He ii (see Section 4.3).
• 6.  
  We include L_bol, M_BH, and L_bol/L_Edd information for EHVO quasars also found in the the SDSS DR7Q sample. The L_bol and M_BH values for the 21 resulting EHVO quasars are overall larger than for the parent sample (2883 quasars) and for BALQSOs (185). L_bol/L_Edd values for EHVO quasars are similar to those of BALQSOs (see Sections 4.4 and 4.4.1).
• 7.  
  We find that it is very unlikely that the distribution of L_bol and M_BH values for EHVO quasars could be obtained randomly from the parent and BALQSO samples (see Section 4.4.1). While little is known about the reddening and bolometric corrections of EHVO quasars, values of M_i[z = 2] for the whole sample support these results (see Section 4.5).
• 8.  
  We observe a trend toward larger M_BH values as the v_min and v_max values of the absorption in BALQSOs increase toward EHVO values. We do not observe a trend for L_bol. There are no trends with BI_EHVO or absorption depth (see Section 4.4.1).
• 9.  
  EHVO quasars show a large range of values of z_em but are overrepresented relative to BALQSOs at larger z_em. The median z_em for EHVO quasars is 2.80, for the parent sample it is 2.42, and for BALQSOs it is 2.39. EHVO quasars represent a large fraction of the SDSS parent quasars at redshifts larger than 4 (see Section 4.5).
• 10.  
  The percentage of radio-loud EHVO quasars is 2.5% (1/40). The only case with a nonzero value of f_ν(20 cm) in FIRST shows a small value f_ν(20 cm) = 1.08 mJy (see Section 4.6). Higher resolution might be needed to determine the true radio nature of EHVO quasars.

The answer is probably both. In Sections 4.5 and 4.6, we showed that, within the current EHVO sample (40 in DR9Q, 21 with information in DR7), EHVO quasars show different distributions from and larger median values for L_bol, M_i[z = 2], and M_BH than both BALQSOs and the quasar parent sample overall. Thus, because they typically are found among intrinsically luminous and massive quasars, EHVO quasars appear to require particular physical properties of the central engine for their outflows to be observed predominantly at these large speeds.

We also have found that the distribution of M_BH values shifts toward larger values as outflow speeds of BALQSOs and EHVO quasars increase. In fact, EHVO quasars and BALQSOs with the largest speeds (${v}_{\min }\,\gt$ 10,000 km s⁻¹ and v_max > 20,000 km s⁻¹) are not significantly distinct in their M_BH values, but they are in L_bol and M_i[z = 2]. BALQSOs overall are on the lower half of L_bol values (Figure 6). In summary, BALQSOs with the largest velocities show similar black hole masses to EHVO quasars, but the luminosities of such BALQSOs are lower than EHVOs; luminosity appears to be the parameter that distinguishes these two classes of quasars.

It is not entirely surprising, given this, that Eddington ratios are not distinct between the three populations. Larger luminosities with similarly larger black hole masses, as EHVO quasars show, and smaller luminosities with smaller black hole masses will result in similar Eddington ratios. We find that EHVO quasars show a large range of Eddington ratios, as do BALQSOs and the quasar parent sample, with the majority of values below the Eddington limit (Figure 8). In a recent study, Yi et al. (2020) did not find significant differences in Eddington ratios for high-z (3 ≤ z_em ≤ 5) BALQSOs compared to non-BALQSOs of the same luminosity and redshift.

Our finding that EHVO quasars may show larger L_bol values overall is associated with the fact that they are predominantly found at larger z_em. We find that the distribution of EHVO quasars is shifted toward larger redshifts relative to our parent DR9Q sample, which ranges from 1.9 < z_em < 4.7 (see Figure 12). EHVO quasars represent 20%–30% of the parent sample for certain redshift bins at z_em > 4. Beyond our sample, the number of known quasars decreases rapidly for z_em > 5: known quasars with redshifts larger than 6 are in the hundreds (Bañados et al. 2016), and only three quasars are reported to date with z_em > 7 (ULAS J1120+0641 in Mortlock et al. 2011, ULAS J1342+0928 in Bañados et al. 2018, and DELS J0038−1527 in Wang et al. 2018). The spectrum of DELS J0038−1527 clearly shows several outflows at extremely high velocities (see Figure 1 in Wang et al. 2018) with v_max ∼ 52,000 km s⁻¹. The spectrum of ULAS J1342+0928 might include EHVO absorption as well, although Bañados et al. (2018) interpreted this absorption as the Gunn–Peterson damping wing of the Lyα. The presence of strong, EHVOs in one of the only three quasars known to date with z_em > 7 is a strong indication that EHVOs might be more common at larger redshifts and a fundamental component of quasars with large L_bol: DELS J0038−1527 is the most luminous quasar with z > 7 (M₁₄₅₀ = −27.1) and Wang et al. (2018) estimate that DELS J0038−1527 has a bolometric luminosity of log (L_bol) = 47.33, similar to the median value of the EHVO quasars in our sample. The black hole mass estimate for DELS J0038−1527, log (M_BH) = 9.12, is, however, lower than the black hole mass estimate of most DR9Q quasars.

At low redshifts z < 1.9, extensive searches for EHVO quasars have not been conducted, and in fact, very few C iv BALQSOs are known (e.g., Ganguly et al. 2007; Dunn et al. 2008; Leighly et al. 2009; Allen et al. 2011). In V. Khatu et al. (2020, in preparation), we show that in a blind sample of 27 AGNs, obtained from HST Cosmic Origins Spectrograph archival data and with redshifts z_em < 0.2, we found no evidence of EHVOs. Hamann et al. (2018) reported a potential C iv outflow with speed ∼0.3c in the spectrum of PDS 456; that was the only case we had labeled as ambiguous due to the presence of unidentified absorption blueward of the Lyα emission line, but there is a lack of another confirming ionic transition in the same outflow at the moment. PDS 456 shows an ultrafast outflow in its X-ray spectrum (see below). The population of EHVO quasars with 0.2 < z_em < 1.9 remains unexplored.

BALQSOs appear typically redder than non-BALQSOs, which has been attributed to being observed at lower inclinations relative to the quasar accretion disk and/or to existing in dustier environments (Weymann et al. 1991; Brotherton et al. 2001; Reichard et al. 2003). In Section 4.5, using Shen et al. (2011) values for the power law of the fitted continuum around the C iv emission line, we observe less negative overall values for BALQSOs than for the quasar parent sample, as expected. EHVO quasars also differ statistically from the parent sample but not from BALQSOs. While this result is preliminary, as the power-law indexes are defined over a small wavelength region (see our discussion in Section 4.5), they might be indicative of EHVO quasars also being observed at lower inclinations and/or in dustier environments, as BALQSOs are. A future study of the values of α_UV (such it was carried out in Baskin et al. 2015) will provide better information of the potential reddening in our sources.

EHVO quasars are not the only outflows detected with very large speeds. UFOs have been observed as Fe K-shell absorption in the X-ray spectra of nearby AGNs (predominantly Seyferts) at speeds similar to or higher (0.03c–0.4c) than the EHVOs in our sample (e.g., Chartas et al. 2002; Tombesi et al. 2010; and references therein). These detections show predominantly narrow and weak absorption features (EW ≤ 150 eV). Given that UFOs have been searched for mostly in local AGNs (z_em < 0.1), the central engine shows typically moderate luminosities ($\mathrm{log}({L}_{\mathrm{bol}})\,\sim$ 43–45.5 erg s⁻¹; this was calculated as ${L}_{\mathrm{bol}}={k}_{\mathrm{bol}}{L}_{\mathrm{ion}}$ assuming a k_bol of 10; see Tombesi et al. 2013). If the trend between velocity and L_bol is extended to the X-ray absorbers, it might imply that UFOs in quasars with larger luminosities would show even larger velocities. The detection of UFOs in high-z quasars is still rare (there are seven cases with z_em > 1.5 at the writing of this paper: the first one being Chartas et al. 2002; see Dadina et al. 2018 and references therein), and mostly in lensed quasars, which make it difficult to constrain their bolometric luminosities (Chartas et al. 2014). A systematic study of UFOs at large redshifts is prohibitively difficult at the moment.

The distinctly higher luminosities of EHVO quasars versus the typically lower luminosities of BALQSOs seems to favor a model where radiation pressure (e.g., Arav et al. 1994; Murray et al. 1995) is responsible for the largest speeds of the outflows we study in this paper. Laor & Brandt (2002) showed that, for soft X-ray-weak quasars (${\alpha }_{\mathrm{OX}}\leqslant -2$), there was a correlation between the maximum velocity of the outflow (v_max) and their luminosity (M_V). Our results seem to initially support this finding. While we do not have direct X-ray information of the EHVO quasars, most EHVO quasars show no signs of He ii λ1640.42 emission, and only one case shows narrow and strong He ii emission (see Section 4.3). The lack of He ii emission is an indicator of softer SED, in which the emission of higher-energy photons is weakened; BALQSOs, for example, have typically weaker He ii emission (Richards et al. 2011; Baskin et al. 2015). In fact, Richards et al. (2011) predicted that quasars with strong He ii emission will be lacking absorption in their spectra; our results agree with this prediction. Surprisingly, though, the EHVO quasar with the largest BI in our sample (J164653.72+243942.2; BI = 3600 km s⁻¹) also shows the strongest He ii emission (see Figure 5). We have already commented on this interesting case in several sections, and we will present a detailed analysis of it in Rodríguez Hidalgo et al. (2020, in preparation). Except for this special case, most EHVO quasars show larger luminosities and potentially softer SEDs than is typical of other quasars.

We find that the three populations of quasars (parent, BALQSO, and EHVO) show a large range of Eddington ratios, and the three populations are indistinguishable in this regard. Large Eddington ratios are, thus, not a requirement to drive these EHVOs.

As discussed in the previous section, UFOs observed in X-ray spectra have similar speeds to EHVOs, but they are typically observed in much less luminous quasars. If there are two different scaling relationships between the outflow terminal velocity and AGN luminosity for UFOs and EHVOs, this might mean that these outflows require different launching or acceleration conditions or that they are entirely different phenomena. One of the differences already found is that UFOs are observed both in radio-quiet and radio-loud AGNs, while EHVO quasars are typically all radio-quiet at FIRST sensitivities (see Section 4.6). We would need to be able to detect UFOs at z_em ∼ 2–5 to be able to compare them appropriately to EHVO quasars.

While the larger luminosities in EHVO quasars suggest that radiation is the driving mechanism, we cannot reject other possibilities such as magnetic driving (de Kool & Begelman 1995; Proga & Kallman 2004; Everett 2005). Indeed, higher terminal velocities are expected in magnetically driven outflows, due to stronger centrifugal forces (Proga 2007). Comparisons to state-of-the-art magnetohydrodynamical simulations obtained with the typical luminosities and black hole masses observed in EHVO quasars will show whether these outflows can be reproduced by the models.

Given the rate of EHVO quasars within the parent population, it is not surprising that we find a lack of EHVO quasars with $\mathrm{log}({M}_{\mathrm{BH}})\lt 9$: there might not be enough quasars per M_BH bin in DR9Q in these ranges to find secure cases of EHVO absorption. However, we do find larger proportions of EHVO quasars at large values of all L_bol, M_i[z = 2], M_BH, and z_em, where the DR9Q quasars are also scarce. Moreover, as discussed before, we find a progression in M_BH values as the velocities of BALs increase toward EHVO values. Once we expand the sample of EHVO quasars, it will be interesting to study this progression with velocity within EHVO quasars. The increase of velocity with M_BH might be an indication of observing larger Keplerian velocities, indicating that the outflows are close to the AGN and the SMBH. However, the distance between these outflows and the AGN source currently is unknown.

Overall, within the current sample, our major finding is the combination of EHVO occurring in quasars with larger luminosities and softer far-UV/X-ray SEDs (as inferred from He ii emission-line strength), which potentially results in outflows easier to drive radiatively. Larger samples of EHVO quasars will allow us to study further these trends, and comparisons to theoretical simulations will help understand the physical properties that drive these outflows.

Assuming all other conditions are equal, faster outflows are more energetic outflows. Outflows have been invoked as one of the potential regulating mechanisms to help establish the relationship found between the black hole masses in the central AGN region and the mass of the surrounding host galaxy (e.g., Silk & Rees 1998; Di Matteo et al. 2005; Springel et al. 2005; Hopkins et al. 2006). In order to determine whether these outflows could significantly affect the host galaxy around them, we need to calculate their kinetic luminosity.

The kinetic or mechanical power can be written as

$$\begin{eqnarray}&&\dot{{E}_{k}}=\displaystyle \frac{1}{2}{\dot{M}}_{\mathrm{out}}{\left({v}_{\mathrm{out}}\right)}^{2},\end{eqnarray} \tag{ 2 }$$

where the mass-outflow rate (${\dot{M}}_{\mathrm{out}}$) is proportional to the velocity (v_out), column density (N_H), distance (r), and, depending on the geometry, some properties such as the global covering factor of the absorber (f_c). Thus, the kinetic power is largely dependent on the velocity of the outflow, as it is proportional to ${({v}_{\mathrm{out}})}^{3}$. Comparing EHVOs to typical BAL outflows results in an increase of 1–2.5 orders of magnitude in kinetic power due solely to this velocity factor. To compare the total kinetic power, we would need good estimates of all of the other parameters (N_H, r and f_c). None of these parameters are yet constrained for EHVO quasars. However, we can set comparisons to other outflow classes by providing some rough estimates: N_H values are likely to be smaller for EHVOs than for strong BAL outflows given that (1) the EHVO absorption does not reach large depths (see Table 2) and (2) the potential low Si iv/C iv ratios suggest that EHVO absorbing gas does not have extremely high optical depths. However, the difference might not be as large if the smaller depths are due to the smaller covering fraction of the absorber. Small covering fractions might be indicative of small distances between the AGN source and the EHVOs. If we assume that each of the other parameters is one order of magnitude smaller than in the case of BAL outflows, the overall kinetic power of EHVO quasars would be similar to that for BALQSOs. However, if EHVOs are at similar distances and have other similar physical and geometrical properties, their kinetic power would be much larger than that of BALQSOs.

Larger samples of EHVO quasars will help us study trends between the outflow velocity and other quasar properties. Detailed studies of individual cases will help constrain the parameters mentioned above and, therefore, the values of mass-outflow rates and kinetic luminosity powers for EHVO quasars. This and future samples of EHVOs will become part of a database that will allow follow-up observations that will help us test physical models of acceleration mechanisms of these outflows.