[OIII] emission in z=2 quasars with and without Broad Absorption Lines

Understanding the links between different phases of outflows from active galactic nuclei is a key goal in extragalactic astrophysics. Here we compare [OIII] $\lambda\lambda$4960,5008 outflow signatures in quasars with and without Broad Absorption Lines (BALs), aiming to test how the broad absorption troughs seen in the rest-frame ultraviolet are linked to the narrow line region outflows seen in the rest-frame optical. We present new near-infrared spectra from Magellan/FIRE which cover [OIII] in 12 quasars with 2.1<z<2.3, selected to have strong outflow signatures in CIV $\lambda$1550. Combining with data from the literature, we build a sample of 73 BAL, 115 miniBAL and 125 non-BAL QSOs with 1.5<z<2.6. The strength and velocity width of [OIII] correlate strongly with the CIV emission properties, but no significant difference is seen in the [OIII] emission-line properties between the BALs, non-BALs and miniBALs once the dependence on CIV emission is taken into account. A weak correlation is observed between the velocities of CIV BALs and [OIII] emission, which is accounted for by the fact that both outflow signatures correlate with the underlying CIV emission properties. Our results add to the growing evidence that BALs and non-BALs are drawn from the same parent population and are consistent with a scenario wherein BAL troughs are intermittent tracers of persistent quasar outflows, with a part of such outflow becoming optically thick along our line-of-sight for sporadic periods of time within which BALs are observed.


INTRODUCTION
Quasar-driven outflows are widely invoked in galaxy formation models in order to reproduce the observed properties of massive galaxies (e.g.Silk & Rees 1998;Springel et al. 2005;Bower et al. 2006;Harrison 2017).Luminous quasars are powerful sources of radiation, and if outflows from the active nucleus can propagate to galaxy scales then the energy in such an outflow would be enough to disrupt the interstellar medium of the host galaxy, preventing star formation and providing an explanation for the observed 'co-evolution' between supermassive black holes and their hosts (Magorrian et al. 1998;Kormendy & Ho 2013).
High-velocity outflows have long been known to exist in many luminous quasars (King & Pounds 2015).Outflow velocities of many thousands of km s −1 are common and are believed to originate from material in a wide-angle outflowing disc-wind (Murray et al. 1995;Giustini & Proga 2019).Such disc-winds are now used to explain the ★ E-mail: Matthew.J.Temple@durham.ac.uk blue-asymmetric profiles of the high-ionization C iv 1550 emission line (Richards et al. 2011;Matthews et al. 2020Matthews et al. , 2023a;;Stepney et al. 2023;Temple et al. 2023;Gillette & Hamann 2024).Up to 50 per cent of quasars exhibit strong, blueshifted absorption due to outflowing material present directly along the line of sight (Weymann et al. 1991;Hall et al. 2002;Allen et al. 2011;Rankine et al. 2020;Bischetti et al. 2023), and the outflow speeds in such 'broad absorption line' (BAL) quasars can exceed 50,000 km s −1 (Bruni et al. 2019;Rodríguez Hidalgo et al. 2020;Rodríguez Hidalgo & Rankine 2022).
Opinions differ as to whether the observed presence of BAL troughs represents a particular evolutionary phase in the quasar fuelling and outflow life-cycle, or a special viewing angle, or if instead BALs are a short, intermittent phase which all quasars will go through stochastically.To address this question, it is informative to consider complementary tracers of Active Galactic Nuclei (AGN) winds which probe different phases and different locations in the outflow (Fiore et al. 2017).One popular probe of ionized gas kinematics in distant galaxies is the [O iii] 4960,5008 1 emission doublet, which is usually inferred to originate in the 'narrow line region' on scales of up to ∼kilo-parsecs (Baskin & Laor 2005;Dempsey & Zakamska 2018).Over recent years, many authors have studied the [O iii] emission properties across different sub-classes of AGN, finding that more luminous quasars generally show weaker [O iii] emission which is broader and often blueshifted, suggesting that [O iii] in luminous quasars is tracing AGN outflows in low-density ionized gas on scales which may be important for host galaxy feedback (e.g.Marziani et al. 2009;Liu et al. 2013;Shen & Ho 2014;Zakamska & Greene 2014;Harrison et al. 2014Harrison et al. , 2016;;Shen 2016;Bischetti et al. 2017;Marziani et al. 2017;Temple et al. 2019;Kakkad et al. 2020;Villar Martín et al. 2020).
However, only a handful of studies have attempted to link quasar outflow signatures that are potentially probing different physical scales (e.g.Elvis 2000;Zamanov et al. 2002;Bruni et al. 2019;Xu et al. 2020;Yi et al. 2020).A holistic understanding of outflow properties probed using different diagnostics is the only way to fully understand the effect AGN feedback has on galaxy formation.Using a sample of ∼200 luminous non-BAL quasars, Coatman et al. (2019) found a correlation between the outflow kinematics of the rest-frame ultraviolet C iv 1550 emission produced on ≲ parsec scales (Fian et al. 2023;Hutsemékers et al. 2023;Shen et al. 2024) and the kinematics of the rest-frame optical [O iii] 4960,5008 emission believed to originate on much larger scales.This result has also been found in smaller samples selected via X-rays (Vietri et al. 2020) and via their high luminosities in either the WISE mid-infrared bands (Vietri et al. 2018) or in optical photometry (Deconto-Machado et al. 2023).Most importantly, these correlations are still seen even when the dependence of both C iv and [O iii] on the quasar luminosity has been taken into account.These results are consistent with a scenario wherein nuclear outflows traced by the C iv emission are capable of propagating to galaxy-wide scales, where they would be able to return a significant amount of energy to the interstellar medium of their host galaxies.
To better understand how BAL outflows are linked to emissionline blueshifts, Rankine et al. (2020) measured the C iv emission line parameters and BAL properties for ≃140 000 quasar spectra from the Sloan Digital Sky Survey (SDSS) DR14 quasar catalogue (Pâris et al. 2018).Spectrum reconstructions based on an independent component analysis (ICA) of a sample of non-BAL quasars allowed for the intrinsic C iv emission of both the BAL and non-BAL quasars to be robustly reconstructed and measured, even in the presence of extensive absorption.Rankine et al. (2020) found that the C iv emission properties of the BAL and non-BAL quasar populations were extremely similar, suggesting that BAL and non-BAL quasars represent different views of the same underlying quasar population.Additionally, BAL trough properties such as the maximum and minimum absorption velocities and the BALnicity index (a measure of the amount of absorption; see Section 2.1 for the definition) were found to strongly correlate with C iv emission line properties.
To further test the hypothesis put forward by Rankine et al. (2020), viz.that BALs and non-BALs are drawn from the same parent population, in this paper we investigate the narrow line region [O iii] emission in a large sample of 73 BAL quasars and compare with the non-BAL population to test whether BALs show evidence for being in a special evolutionary phase.More precisely, in a model where the BAL outflows represent a specific phase in the quasar fu- elling/outflow cycle, the extended narrow line region emission would be expected to differ significantly between the BAL and non-BAL quasars (Turnshek et al. 1997).The impact of the energetic BAL flows would reduce the emission from the static narrow line region although broader, more blueshifted [O iii] emission may be seen (e.g.Zakamska et al. 2016).At a given C iv emission-line blueshift and equivalent width (EW), [O iii] in the BAL quasars would have lower EW and be more blueshifted compared to the non-BALs.Alternatively, if BALs are stochastic phenomena which may appear intermittently for short periods of time while any galaxy is in a luminous quasar phase, then their [O iii] properties should be very similar to non-BALs.
Previous near-infrared observations of  > 1 quasars have mostly focused on non-BAL quasars (e.g.Coatman et al. 2019), and so in Section 2 we present 12 new Magellan/FIRE spectra which were specifically targeted to observe the rest-frame optical [O iii] in BAL quasars.We combine with archival data from the literature to build a sample of 73 BAL, 115 miniBAL and 125 non-BAL quasars, as defined in Section 2.1.In Section 3 we then measure the [O iii] strength and kinematics in our sample of BAL quasars, and compare the [O iii] outflow signatures seen in the BAL, miniBAL and non-BAL populations.We discuss our findings in Section 4 and summarise our conclusions in Section 5.

SAMPLE AND DATA
To investigate the link between the outflows seen in the rest-frame ultraviolet and rest-frame optical wavebands, we compile a sample of 1.5 <  < 2.6 quasars with both SDSS observed-frame optical spectra (subsection 2.1) and near-infrared spectra from various sources (subsections 2.2 and 2.3).To allow identification of broad absorption features up to 25 000 km s −1 bluewards of C iv, we require redshift  > 1.56 for objects with spectra from BOSS and  > 1.67 for objects with data from the original SDSS spectrograph (i.e.observations before the 'SDSS MJD' of 55000).The  < 2.6 criterion allows us to check for low-ionization Mg ii broad absorption troughs in our final sample; no such 'LoBALs' are found.To ensure reliable detection of absorption troughs in the SDSS spectra we require the average signal-to-noise ratio per pixel to be ≥ 10, as the BAL fraction has been found to depend only weakly on the signal-to-noise ratio above this threshold (fig. 2 of Rankine et al. 2020, see also fig. 4 of Gibson et al. 2009).To ensure coverage of [O iii], H and H in the infrared JHK bands, we require either 1.56 <  < 1.65 or 1.95 <  < 2.60, to avoid emission lines falling in the regions of low atmospheric transparency between the JHK bands.
We measure the rest-frame ultraviolet monochromatic luminosity   at  = 3000 Å (hereafter  3000 ) by fitting a model spectral energy distribution (Temple et al. 2021b) to the griz SDSS photometric data.Our final sample spans 8 × 10 45 <  3000 < 2 × 10 47 erg s −1 .The bolometric correction for each object is likely in the range  bol / 3000 ≈ 3-10 ( Temple et al. 2023, fig. 6), so all of the objects in our sample lie well above the  bol ≳ 3 × 10 45 erg s −1 threshold which was suggested by Zakamska et al. (2016) as necessary for [O iii] winds to contribute to quasar feedback.
Our compilation results in a total of 313 unique SDSS quasars with optical and near-infrared spectra covering the rest-frame 1400-2800 and 4800-6600 Å wavelength ranges, as shown in Figs. 1 and  2. The majority of our sample (276/313) have 1.95 <  < 2.6; only 37 quasars have 1.56 <  < 1.65.We have verified that the results and conclusions of this paper would not change if we restricted our sample to 2 <  < 2.5 and 10 46 <  3000 < 10 47 erg s −1 .From our sample of 313 quasars, 125 objects show no C iv absorption exceeding ≈ 450 km s −1 in the rest-frame ultraviolet ('non-BALs'), 115 show mild absorption with trough widths > 450 km s −1 which does not meet the strict definition of a BAL ('miniBAL' quasars), while 73 sources are bona fide BALQSOs according to the definition of Weymann et al. (1991).All of our BALQSOs are so-called 'Hi-BAL' systems which show no evidence for low-ionization absorption troughs.
We now describe in detail the data sets used, including the SDSS optical data (subsection 2.1), 12 new near-infrared observations from the FIRE spectrograph (2.2) and existing near-infrared spectra from the literature (2.3).In Section 3 we then describe the methods used to analyse these spectra.

SDSS spectra and C iv measurements
We use the same C iv emission-line information as described in section 2.1 of Temple et al. (2023).In brief, we start with all quasars from the 16th and 17th data releases of SDSS (Lyke et al. 2020;Abdurro'uf et al. 2022).For this work, we consider only quasar spectra with mean signal-to-noise ratio (per 69 km s −1 pixel) ≥ 10 over the rest-frame interval 1700-2200 Å.Each spectrum is reconstructed using the independent component analysis (ICA) scheme described by Rankine et al. (2020).A linear combination of ten spectral ICA components is used to model the data, using an iterative routine to mask absorption features.This approach allows the underlying emission-line properties to be inferred consistently in both the BAL and non-BAL quasars.Examples of the SDSS data and corresponding ICA reconstructions are shown in Appendix A. We keep only ICA reconstructions with reduced- 2 < 2. The C iv emission-line blueshift is measured as the Doppler shift of the median continuumsubtracted line flux, assuming a rest-frame wavelength of 1549.48Å.
The uncertainty on the C iv blueshift is dominated by the systemic redshift uncertainty, which is ≲ 250 km s −1 for our  ≈ 2 quasars (Hewett & Wild 2010).The uncertainties associated with our C iv emission-line EWs are dominated by the time variability of individual quasars, as for example investigated by Rivera et al. (2020) using the SDSS-RM sample of high-cadence repeat spectroscopic observations.From fig. 9 of Rivera et al. (2020) we see that the C iv EW can vary by up to 20 per cent as a function of observation epoch.
BAL quasars are identified via the BALnicity Index (BI; Weymann et al. 1991) and the Absorption Index (AI; Hall et al. 2002).The former is defined as with  = 1 when  () < 0.9 contiguously for at least 2000 km s −1 . () < 0.9 is the continuum-normalised spectrum; in this case normalised by the reconstruction.The lower integration limit of 3000 km s −1 is set to remove any contribution of strong 'associated absorbers' while the upper 25 000 km s −1 limit avoids confusion with absorption of the Si iv 1397 ion.The absorption index, on the other hand, includes absorption below 3000 km s −1 and requires  () < 0.9 contiguously for at least only 450 km s −1 : There exists a bimodal distribution of log(AI) first observed by Knigge et al. (2008): one population of quasars have both AI>0 and BI>0, where, for the majority, the AI and BI are measuring the same absorption trough(s).The second population have AI>0 but BI=0 due to the presence of only narrow troughs, or broad troughs where a significant fraction of the absorption occurs below 3000 km s −1 .Here we use 'miniBAL' to refer to the second AI>0 population with BI=0 i.e. quasars which have absorption troughs wider than 450 km s −1 without meeting the Weymann et al. (1991) (Skrutskie et al. 2006) photometry with either  < 17.1 or ( + )/2 < 17.1 to ensure good signal-to-noise ratio in the resulting near-infrared spectra.Targets were then prioritised based on their C iv properties: objects with large C iv blueshifts were preferred, as these are under-represented in existing samples.Targets were observed on the nights of 2022 January 02 and 03 using FIRE in echelle mode with the 0.6 arcsec slit, which delivers  = 6000 spectra across 0.82-2.51m.Seeing was in the range 0.5 to 0.9 arcsec with dark moon and no clouds.12 science targets were observed, including seven BALs and three non-BALs with CIV blueshift > 2000 km s −1 and two filler targets (one BAL and one non-BAL) with shorter exposures, as summarised in Table 1.Spectra were reduced using the standard set-up in PypeIt v1.11 (Prochaska et al. 2020;Westfall 2022).Targets were nodded along the slit in ABBA dither pattern and combined in AA-BB groups for sky subtraction.Wavelength calibration was performed using night-sky emission features.Bright A0V stars observed directly before or after each science target were used for flux calibration.Spectra were then corrected for telluric absorption using a model grid.We show the reduced JHK spectra in Fig. 3: some residual telluric features can be observed around 20000-20150 Å, but H and H are clearly detected in each object.[O iii] is clearly seen in e.g.J100711+053208, but in many objects there is strong Fe ii emission in the H band and spectral modelling is required to deblend H, [O iii] and Fe ii.

Existing near-infrared spectra
To complement our targeted observations with Magellan/FIRE, we also cross-match our parent sample of SDSS quasars to various catalogues from the literature to build a large sample of objects with coverage of C iv and [O iii].Starting from the SDSS sample described in Section 2.1, we look for near-infrared data from the GNIRS 'Distant Quasar Survey' (Matthews et al. 2021(Matthews et al. , 2023b)), the previous compilation of Coatman et al. (2017Coatman et al. ( , 2019)), the X-Shooter programme described by Xu et al. (2019Xu et al. ( , 2020)), and the SUPER survey (Circosta et al. 2018).For objects which have more than one near-infrared spectrum available, we keep only the spectrum with the highest signal-to-noise ratio in the 4800-5100 Å region which contains H and [O iii].

GNIRS data from the DQS
The Gemini/GNIRS 'Distant Quasar Survey' (DQS; PI: Shemmer; Matthews et al. 2021Matthews et al. , 2023b) is a large and long programme to obtain high signal-to-noise ratio near-infrared spectra for a large sample of SDSS quasars at 1.5 <  < 3.5.Raw data were downloaded from the Gemini archive and reduced using PypeIt.190 objects from the GNIRS-DQS match to our sample of SDSS spectra.

Coatman et al. compilation
The largest previous catalogue of [O iii] emission from near-infrared quasar spectra was described by Coatman et al. (2019).This compilation includes near-infrared spectra from WHT/LIRIS observations presented by Coatman et al. (2016), TRIPLESPEC and FIRE observations presented by Shen & Liu (2012) and Shen (2016), observations from the 'Quasars probing Quasars' project (Hennawi et al. 2006), as well as programmes from P200/TRIPLESPEC, VLT/SINFONI and NTT/SOFI.The construction of this data set is described in full by Coatman et al. (2017).For the purposes of this investigation, we keep only spectra with median signal-to-noise ratio (per 69 km/s pixel) > 3 across the rest-frame 4800-5100 Å region.One-hundred spectra from this catalogue with coverage of H, [O iii] and H are matched to our SDSS sample.2020) presented near-infrared observations of five BAL and two miniBAL quasars.These data were taken from a wider VLT/X-Shooter programme (PI: Benn) which observed a total of 20 quasars with rest-frame ultraviolet absorption features (Xu et al. 2019).Ten of these X-Shooter sources match to our SDSS parent sample.For each source, we downloaded the Phase 3 data products from the ESO archive.Each exposure was corrected for telluric absorption using molecfit (Kausch et al. 2015;Smette et al. 2015).Where more than one exposure was present for a source, all such observations were combined to give the best signal-to-noise ratio spectrum for each quasar.

SINFONI data from the SUPER survey
The SUPER survey (Circosta et al. 2018;Kakkad et al. 2020;Vietri et al. 2020) is a VLT/SINFONI large programme (PI: Mainieri) designed to study AGN feedback at so-called 'cosmic noon',  ≈ 2. Targets were selected from X-ray surveys to have redshifts 2 <  < 2.5.We downloaded the DR1 data products from the ESO archive, which consist of the combined, flux-calibrated data cubes for 20 Type-1 AGN presented by Kakkad et al. (2020).We extract 1d spectra from the SUPER data cubes using 0.6 arcsec apertures to match the slit width of our FIRE observations, and estimate 1d noise arrays from the pixel-pixel variations away from the source in each data cube.We find that one SUPER source with SINFONI H+K data covering H, [O iii] and H, has SDSS data which meet the criteria described in Section 2.1.

Spectral modelling procedure
We use the open source code fantasy2 to simultaneously model the [O iii], Fe ii, and Balmer emission lines in each observed-frame optical spectrum from our sample of quasars.fantasy is a python code for simultaneous multi-component fitting of AGN spectra, described by Ilić et al. (2020Ilić et al. ( , 2023) ) and Rakić (2022).For our luminous (10 46 ≲  bol ≲ 10 48 erg s −1 ) quasars, we do not include a host galaxy component.The Balmer lines (H and H, and where available H and H) are kinematically tied to have identical velocity profiles, with up to two broad and one narrow Gaussian components, although the relative normalisation of each line is free to vary.A comprehensive set of Fe ii emission blends are included as described in Ilić et al. (2023).We require coverage of H to ensure robust modelling of the red wing of H in objects with weak [O iii] and strong Fe ii blended around 4900-5100 Å.
We fit each spectrum twice: once with no [O iii], and once with one broad and one narrow [O iii] component.The 4960 and 5008 Å lines are constrained to have identical kinematics with the amplitudes tied in a 1 : 3 ratio.The Bayesian Information Criterion (BIC) is calculated for each fit.For quasars where the BIC is not improved by 10 or more when including [O iii], we flag the [O iii] component as not robustly detected, and exclude the spectrum from our kinematic analysis.Many of these spectra would be inferred to have extremely weak (rest-frame) [O iii] EW < 1 Å.Two-hundred and twenty-eight objects from our sample of 313 quasars are judged to have robust [O iii] detections: 57 BALs, 78 miniBALs and 93 non-BALs.For these objects we measure  80 , the velocity width containing 80 per cent of the total 5008 Å line flux, as is commonly used in the literature to quantify [O iii] outflow signatures (Zakamska & Greene 2014;Coatman et al. 2019;Villar Martín et al. 2020).with previous works (Zakamska & Greene 2014;Coatman et al. 2019;Villar Martín et al. 2020).However, the dynamic range in luminosity spanned by the majority of our sample is relatively small (≈1 dex), and the correlation observed between [O iii] and C iv is not a secondary effect driven by an underlying correlation with the intrinsic quasar luminosity (such as the Baldwin effect).

Our
For our BAL and miniBAL quasar samples, we test if the absorption trough properties are linked to the kinematics of the narrow line region traced by the [O iii] emission.In Fig. 5 we show the [O iii] EW and  80 as a function of absorption trough properties for both the miniBAL and the bona fide BAL populations.We show the Weymann et al. (1991) BALnicity Index (BI) and Hall et al. (2002) Absorption Index (AI), the absorption trough width (or median width, for quasars with more than one trough), and the fastest outflow velocity of the absorption trough (i.e. the fastest velocity contributing to the BI or AI measurements in Equations 1 and 2), as these parameters have been seen to show the strongest correlations with C iv emission line parameters (Rankine et al. 2020).We find a mild but statistically significant correlation between the maximum BAL trough velocity and the [O iii]  80 ( = 0.37,  = 0.004).However, if we express the maximum trough velocity in units of the C iv emission-line blueshift the correlation with [O iii] is not present, suggesting that the BAL is tracing the high-velocity tail of the C iv emission-line outflow,  max ≈ 10 times the median C iv emission blueshift.None of the other BAL trough parameters computed by Rankine et al. (2020), such as the minimum trough velocity and velocity of the deepest part of the trough, is found to correlate with the [O iii] measurements computed in this work.For BALQSOs, we show the BI (Equation 1) and the width and maximum velocity of the BAL trough.For miniBALs, we show the AI (Equation 2) and the width and maximum velocity of the miniBAL trough.For objects with more than one trough, we plot the median trough width and the fastest maximum velocity.Our absorption-finding algorithm only searches up to 25 000 km s −1 , to avoid confusion with Si iv absorption, so the cluster of points on the right-hand edge of the right-hand panels (shown as triangles) could be considered lower limits on  max .

DISCUSSION
The observational results presented in the previous section can be summarised as three key findings.First, larger C iv emission-line blueshift is observed to correlate with weaker C iv EW, weaker [O iii] EW and broader [O iii] velocity width, confirming the correlations found in smaller samples by previous authors (Bachev et al. 2004;Marziani et al. 2017;Vietri et al. 2018;Coatman et al. 2019;Vietri et al. 2020;Deconto-Machado et al. 2023).Second, BAL and miniBAL quasars do not show significant differences in their [O iii] emission compared to their non-BAL counterparts.And third, while there is a mild correlation between the BAL outflow velocity and the [O iii] velocity width, this is most likely driven by the fact that these parameters are both correlated with the underlying C iv emission-line kinematics (Coatman et al. 2019;Rankine et al. 2020).
In this section, we first discuss these findings in the context of previous work, and then explore possible interpretation.

Comparison with previous BAL investigations
We believe our investigation into the rest-frame optical [O iii] properties of 73 BAL quasars represents the largest such sample available to date.Recently Xu et al. (2020) conducted a similar investigation, measuring the [O iii] strength and kinematics in a smaller sample of five BAL and two miniBAL quasars.Xu et al. (2020) selected their sample to have high-ionization Si iv BALs in addition to C iv troughs, allowing them to infer constraints on the density and radius of the absorbing gas.They find that the electron number density derived for the absorbing material increases with decreasing [O iii] EW. 3 They also find that the measured velocity widths have similar sizes in the BAL troughs and [O iii] emission lines, and suggest that this is consistent with the [O iii] emission and BAL absorption being "different manifestations of the same wind".Our [O iii] velocity widths span a similar range to those found by Xu et al. (2020), with  80 in the range 1000-2500 km s −1 and  90 in the range 1000-3000 km s −1 .However, our BAL trough widths span a larger range, up to more than 20,000 km s −1 , and with our larger sample we do not find a significant correlation between the BAL trough widths and the [O iii] velocity widths.Unlike Xu et al. (2020), we therefore do not believe that the [O iii] emission and BAL absorption need to be tracing the same gas in each object, that is to say outflowing gas with the same density, ionization parameter and location.
As this paper was going through the peer review process, Ahmed et al. (2024) was posted on the arXiv.These authors investigated 65 BAL quasars from the GNIRS-DQS and found no significant differences in their rest-frame optical spectra compared to a control sample of non-BAL quasars.Our results are in agreement with Ahmed et al. (2024), although we note that our larger sample size and new Magellan/FIRE data together also allow us to investigate the [O iii] emission in a statistical sample of BALs with more extreme C iv blueshifts ≳ 2500 km s −1 .
In this work we have focused on high-ionization BAL quasars, which do not show absorption troughs from low-ionization lines such as Mg ii 2800 and Al iii 1860.Quasars with such low-ionization BAL troughs ('LoBALs') can be identified at lower redshifts when Mg ii or Al iii is present in the SDSS observed-frame optical spectrum (Voit et al. 1993).Schulze et al. (2017) presented near-infrared observations of 22 LoBALs at 1.3 <  < 2.5, finding no enhancement in [O iii] outflow signature compared to the wider non-BAL population.Matthews et al. (2017) also found no difference in distribution of [O iii] EWs in 58 LoBALs at 0.35 <  < 0.83, when compared to wider SDSS quasar population at the same redshift.Our results are consistent with these works to the extent that neither our high-ionization BAL quasars nor their low-ionization BAL quasars show any significant difference from the non-BAL population in terms of their optical [O iii] emission, suggesting that the narrow line region properties are not significantly impacted by the presence of a broad absorption feature in the rest-frame ultraviolet.

Implications for quasar winds and feedback
This investigation was motivated (in part) by the idea that if BAL and non-BAL quasars are members of two distinct AGN populations with different disc-winds, driving ionized gas outflows with differing powers, then we might expect BALs and non-BALs to show differences in their narrow line region (as traced by the [O iii] strength and kinematics).However, this is true only if the length of the BAL phase is comparable to the time taken to affect the narrow line region.Estimates of quasar lifetimes are typically of the order of 1-10 Myr (Khrykin et al. 2021), which together with the significant (∼10-40 per cent) observed BAL fraction (Allen et al. 2011) means that if the BAL phenomenon was a single evolutionary phase then it would typically last at least ∼ 0.1 Myr.In this time a BAL outflow travelling at 10 000 km s −1 would travel ∼1 kpc, meaning that we would expect it to reach the scales which we are probing with [O iii].In Section 3.2 we found that the BAL, miniBAL and non-BAL quasar populations show no significant differences in their narrow line region [O iii] properties.Our results therefore suggest that the BAL phenomenon is not a distinct, long-lived evolutionary phase in the cycle of SMBH growth, at least within the context of luminous quasar activity.
One possible interpretation could instead be that BAL outflows don't do anything to affect the [O iii]-emitting gas in luminous quasars -either suggesting that the kinetic power contained in BAL outflows is negligible, or that they are entrained in a particular geometry which means that they are directed away from the narrow line region gas.This would imply either that BAL outflows are not important for quasar feedback, which is unlikely given that photoionization models of BAL troughs suggest that at least some BALs carry significant kinetic power (Arav et al. 2018;Xu et al. 2019), or that the [O iii]emitting region is not tracing the impact of quasar-mode outflows on the interstellar medium of the AGN's host galaxy.
On the other hand, if we assume that BAL outflows should affect the narrow line region, then the similarity seen in the BAL and non-BAL quasars would suggest that the observation of a BAL trough is an intermittent, but recurring phenomenon during the luminous quasar phase of the SMBH growth cycle.There is a growing body of work suggesting that BAL troughs can vary on relatively short timescales, which would support this hypothesis (Hamann et al. 2008;Gibson et al. 2008Gibson et al. , 2010;;Capellupo et al. 2011Capellupo et al. , 2013;;Filiz Ak et al. 2012, 2013;Grier et al. 2015;McGraw et al. 2017;De Cicco et al. 2018;Rogerson et al. 2018;Hemler et al. 2019;Mishra et al. 2021;Vietri et al. 2022;Aromal et al. 2023).In other words, our results would be consistent with a scenario where BAL troughs are an intermittent tracer of a persistent quasar outflow (which could be the wind traced by the C iv emission blueshift), where a part of the outflow along our line-of-sight becomes optically thick for short periods of time when a broad absorption trough is observed.This scenario would explain the correlations observed between the C iv and [O iii] outflow signatures: luminous quasars which drive winds on nuclear scales are also able to drive outflows on much larger scales, with stochastic parts of the wind sometimes producing absorption features which don't correlate directly with the properties of the narrow line region outflow.The general properties of such nuclear winds would then be governed by the shape and strength of the ionizing quasar continuum, which is in turn set by the SMBH mass and accretion rate (Temple et al. 2023).This scenario would also explain the fact that BAL and non-BAL quasars have no significant difference in their underlying C iv emission (Rankine et al. 2020) or their sublimation-temperature dust emission (section 4.3 of Temple et al. 2021a, see also Saccheo et al. 2023), but a further element would be required to explain the differences in radio properties of the BAL and non-BAL populations observed by Petley et al. (2022Petley et al. ( , 2024)), especially as the radio emission in AGN is often found to be linked with the [O iii] properties (e.g.Jarvis et al. 2021).

CONCLUSIONS
We have measured the rest-frame optical [O iii] 4960,5008 emission in a sample of 73 BAL, 115 miniBAL and 125 non-BAL quasars at 1.56 <  < 2.6 with high-quality near-infrared spectroscopic data.
Our key observational results are: (ii) BAL, miniBAL and non-BAL quasars show no significant differences in their [O iii] emission: all three sub-samples show the same correlations described in (i).
(iii) In BALQSOs, the maximum absorption trough velocity shows a weak correlation with the [O iii] velocity width, but this is fully explained by the dependence of both quantities on the C iv emission-line blueshift.When the BAL outflow velocities are normalised by the C iv blueshift, we find no correlations between the [O iii] emission kinematics and the C iv BAL trough parameters.
Our results disfavour a scenario where (high-ionization) BALQ-SOs are a special, long-lived evolutionary phase in which more powerful (cf.non-BAL QSOs) winds are able to propagate into the interstellar medium of their host galaxies, and either clear out or alter the kinematics of this medium.Instead our results are consistent with a scenario in which BAL troughs are an intermittent, stochastic phenomenon which all luminous quasars with persistent outflowing winds (traced by both [O iii] and C iv emission) are likely to undergo.This is consistent with observations of BAL trough variability on time-scales of only months to years, and supports the scenario favoured by Rankine et al. (2020) where BALs and non-BALs are members of the same underlying quasar population.

Figure 3 .
Figure3.Magellan FIRE spectra for twelve quasars which are newly presented in this work (Section 2.2), with a resolution of 50 km s −1 ( = 6000).Gray shows the spectra, with a 9-pixel inverse variance weighted smooth in black.The expected wavelengths of H, [O iii] and H are marked in red, assuming the redshifts estimated from the SDSS rest-frame ultraviolet spectra.The corresponding SDSS spectra are shown in Appendix A.
first observational result is the dependence of the [O iii] EW on the C iv emission-line blueshift and EW, as shown in the bottom panels of Fig. 4. For objects with robust [O iii] measurements, we show the  80 velocity width in the top panels of Fig. 4. Consistent with previous works (e.g.Coatman et al. 2019), we find that the [O iii] properties are correlated with the ultraviolet C iv morphology.Objects with larger C iv blueshifts have weaker [O iii] EW (Pearson's  = −0.42, = 10 −14 ) with broader [O iii]  80 ( = 0.45,  = 10 −12 ).Stronger C iv EW is correlated with stronger [O iii] EW ( = 0.65,  = 10 −39 ) and narrower [O iii]  80 ( = −0.34, = 10 −7 ).When we compare our BAL, miniBAL and non-BAL quasar samples, we see no difference in their [O iii] emission, when the underlying dependence on the C iv emission properties is taken into account.In other words, we see no evidence for BAL or miniBAL quasars having different narrow line region properties when compared to their non-BAL counterparts.We also see a weak trend with luminosity: objects with larger  3000 are more likely to show smaller [O iii] EW ( = −0.20, = 0.00037) and broader [O iii]  80 ( = 0.23,  = 0.00048)

Figure 4 .
Figure 4. [O iii] EW and  80 as a function of C iv properties and  3000 .The typical uncertainty associated with each individual data point is shown by the black crosses in the upper right of each panel.Open symbols indicate sources where ΔBIC < 10 when adding [O iii] to the spectral model; these [O iii] lines are not considered to be robustly detected and so we do not attempt to derive velocity width information.Objects with weaker [O iii] are plotted at 1 Å EW for display purposes only.The [O iii] strength and velocity width correlate with the C iv emission line blueshift and EW, with the same trends observed in our samples of BAL, miniBAL and non-BAL quasars.

Figure 5 .
Figure 5. [O iii] EW and velocity width as a function of C iv absorption properties.The typical uncertainty associated with each individual data point is shown by the black crosses in the upper left of each panel.As in Fig. 4, open symbols indicate objects where [O iii] was not robustly detected in the near-infrared spectrum.For BALQSOs, we show the BI (Equation1) and the width and maximum velocity of the BAL trough.For miniBALs, we show the AI (Equation2) and the width and maximum velocity of the miniBAL trough.For objects with more than one trough, we plot the median trough width and the fastest maximum velocity.Our absorption-finding algorithm only searches up to 25 000 km s −1 , to avoid confusion with Si iv absorption, so the cluster of points on the right-hand edge of the right-hand panels (shown as triangles) could be considered lower limits on  max .

3
Xu et al. report  [O iii] / bol , which is closely related to the EW.
(i) The properties of [O iii] 5008 and C iv 1550 emission are connected: larger C iv emission-line blueshift and weaker C iv EW correlate with weaker [O iii] EW and broader [O iii] velocity structure.These correlations are not driven solely by changes in the 3000 Å continuum luminosity.

Table 1 .
Quasars with new near-infrared spectra obtained from Magellan/FIRE, as described in Section 2.2.

Table 2 .
Description of columns for the supplemental data file available at MNRAS online.