BAL Outflow in Quasar B0254-3327B: Analysis and Comparison with Other Extreme UV Outflows

We have identified a broad absorption line (BAL) outflow in the HST/STIS spectrum of the quasar QSO B0254-3327B at velocity $v=-3200\text{ km s$^{-1}$}$. The outflow has absorption troughs from ions such as Ne VIII, Na IX, Si XII, and Ne V. We also report the first detection of S XIV absorption troughs, implying very high ionization. Via measurement of the ionic column densities, photoionization analysis, and determination of the electron number density of the outflow, we found the kinetic luminosity of the outflow system to be up to $\sim1\%$ of the quasar's Eddington luminosity, or $\sim5\%$ of the bolometric luminosity, making it a potential contributor to AGN feedback. A solution with two ionization phases was needed, as a single phase was not sufficient to satisfy the constraints from the measured ionic column densities. We find that the ionization parameter of the very high-ionization phase of the outflow is within the expected range of an X-ray warm absorber. We also examined the physical properties of the outflow of Q0254-334 along with previously studied extreme UV outflows, with a total sample of 24 outflow systems, finding a weak negative correlation between outflow velocity and distance from the central source, with larger distances corresponding to slower velocities. The very high-ionization phase of the Q0254-334 outflow has one of the highest ionization parameters of UV absorption outflows to date, which we attribute to the presence of S XIV.

In order to find the value of   , it is important to find the mass flow rate ( ), a method for which involves finding the electron number density (  ) and ionization parameter (  ) to measure the distance () of the outflow from the central source (Borguet et al. 2012b).Multiple quasar outflows have been analyzed via this method (e.g. de Kool et al. 2001;Hamann et al. 2001;Walker et al. 2022;Byun et al. 2022a).For ionized outflows, the ionization parameter can be deter-★ E-mail:dbyun@vt.edumined by measuring the column densities of ions, and comparing them with simulated values based on a range of   and hydrogen column density (  ).Multiple outflow analysis studies have been conducted using the spectral synthesis code Cloudy (Ferland et al. 2017) for this method.(e.g.Xu et al. 2018;Miller et al. 2020a;Byun et al. 2022a;Walker et al. 2022).This paper presents the analysis of the absorption outflow of the quasar QSO B0254-3327B (hereafter Q0254-334), using the method described above, based on HST/STIS observational data, ultimately finding the ratio between   and   .
The paper is structured as follows.Section 2 describes the observation and data acquisition of Q0254-334; section 3 discusses the process of finding the ionic column densities,   ,   , and   of the outflow; section 4 shows our analysis results of whether the outflow's kinetic luminosity is sufficient to contribute to AGN feedback, and compares our results with studies of other outflows; and section 5 concludes and summarizes the paper.In our analysis, we adopted a cosmology of ℎ = 0.696, Ω  = 0.286, and Ω Λ = 0.714 (Bennett et al. 2014).We used the Python astronomy package Astropy (Astropy Collaboration et al. 2013, 2018) for our cosmological calculations.We also used Scipy (Virtanen et al. 2020), Numpy (Harris et al. 2020), andPandas (v1.2.4, Reback et al. 2021;Wes McKinney 2010) for the majority of our numerical computations, as well as Matplotlib (Hunter 2007) for plotting our figures., with the G230L and G140L gratings respectively.Prior to this, it was also observed with HST/FOS in 1994.Due to the limited wavelength range of the FOS data relative to that of STIS, we have focused on the STIS data for the purpose of this analysis.After retrieving the data from the Mikulski Archive for Space Telescopes, we have co-added the two STIS spectra, and corrected the combined spectrum for galactic reddening and extinction with  ( − ) = 0.0205 (Schlafly & Finkbeiner 2011), and the extinction model by Fitzpatrick (1999).The co-added and dereddened spectrum of the two observations, covering observed wavelengths 1138.6-3156.6Å,along with the FOS spectrum, is shown in Figure 1.

OBSERVATION AND DATA ACQUISITION
We have identified a broad absorption line (BAL) outflow system at  = −3200 km s −1 , with its ionic absorption troughs marked by red vertical lines in Figure 1.Troughs exist of species such as Ne viii, Na ix, and Si xii, as well as excited state transitions such as O iv* and Ne v*.Arav et al. (2020) define a BAL in the extreme UV range as a continuous absorption feature with normalized flux  ≤ 0.9 over a width of Δ ≳ 1500 km s −1 , at least −3000 km s −1 blueward of the center of emission.We have verified that the outflow is a BAL outflow by confirming the width of the Si xii 499.406and Ne viii 780.324troughs (see Figure 2).The normalized flux was found by modeling the quasar's continuum via a spline model that gave the minimum possible continuum above the absorption, and the most prominent emission features (e.g.O vi) with Gaussians.The presence of the Ne v* feature allowed us to find the value of   , as shown in Section 3.2.

Ionic Column Densities
As the ionic column densities (  ) of the outflow are crucial in finding the physical properties of the outflow, we used two different methods to find them based on the absorption troughs: the apparent optical depth (AOD) method in which we assume uniform and homogeneous covering (Savage & Sembach 1991); and the partial covering (PC) method in which we include a covering factor  < 1 (Barlow et al. 1997;Arav et al. 1999b;Arav et al. 1999a).
The AOD method allows us to find upper limits and lower limits of ionic column densities with its relative simplicity, while the PC method lets us find more accurate measurements of ions with doublet features (e.g. de Kool et al. 2002;Arav et al. 2005;Borguet et al. 2012a;Byun et al. 2022b).As done by Byun et al. (2022c) for the quasar J024221.87+004912.6,we selected the appropriate method for computing the column density of each ion.
The AOD method involves the following relation between intensity and optical depth (Spitzer 1978;Savage & Sembach 1991): (1) where  () is the intensity as a function of wavelength,  0 () is the intensity without absorption, and  is the optical depth.Finding the optical depth enables computation of the column density, as they have the following relation: where () is the optical depth as a function of velocity,  is the elementary charge,   is the mass of an electron, and  () is the column density per unit velocity.Integrating  () over the velocity range of an ion's absorption trough results in the ion's column density.
As mentioned above, the PC method involves a covering factor  < 1, which follows the relation shown in the equations below (Arav et al. 2005): (3) where   () and   () are the intensities at the red (longer wavelength) and blue (shorter wavelength) troughs of a doublet transition,  () is the covering factor as a function of velocity, and  is the optical depth.
For each ion, we converted the spectrum from wavelength space to velocity space, using the redshift of the quasar and the wavelengths of the ionic transition lines (see Figures 3,4).We then chose integration ranges for each ion that covered visible absorption features and minimized blending effects with other lines.For instance, the O vi doublet had heavy blending between the red and blue troughs (see Figure 3f).We thus chose a range where the overlap between the red and blue troughs would be minimized and computed a lower limit to the column density with the AOD method.As there were no discernible absorption troughs of Ly , C iii, and S iv*, we measured their AOD column density with integration range  ≈ −4500 to −2000 km s −1 to match the Ne viii width, and treated them as upper limits.Due to the severe blending in the multiplet of S iv 744.904,748.393 and S iv* 750.221(see panel 4 of Figure 1), we were unable to pinpoint the column density of the resonance state S iv 0 from this trough.However, as there was no discernible absorption trough of S iv 809.656,we were able to find an upper limit of its column density.Similarly, the trough of O iv 787.711blended with O iv* 790.190,and potentially with the neighboring S v 786.468.As such, we were unable to find the column density of the resonance state O iv 0, and could only find a lower limit of the O iv* column density.We were also limited to finding a lower limit of the Ne viii column density based on an AOD measurement due to the saturation of the doublet troughs.In the doublet of S xiv, we determined the red trough of S xiv 446 to be contaminated, due to the visible blueward absorption compared to the blue trough 418 (see Figure 4  plot c).Due to this limitation, we measured the AOD column density of S xiv based on the blue trough.We determined that it was safe to treat this column density as a measurement, due to its shallower depth relative to similarly ionized troughs with comparable oscillator strengths (e.g.Si xii).
The integrated column densities are shown in Table 1.The rightmost column shows the values adopted for the photoionization solution described in Section 3.3.The errors have been propagated from the error in the flux, and a conservative 20% error has been added in quadrature to the adopted column density errors to account for the uncertainty in the continuum level due to the subjectivity of the model (Xu et al. 2018).This uncertainty is demonstrated in the column density calculation of O iv* based on the different continuum models shown in Figure 5.The maximum, minimum, and intermediate continuum fits in the region are shown as blue, red, and purple dashed lines respectively.The O iv* absorption is marked with a gray vertical line.The AOD measurement of the O iv* column density is 28.0 +2.8 −2.0 × 10 14 cm −2 for the intermediate continuum, while the higher and lower continuum levels yield results of 32.6 +2.8 −2.0 × 10 14 cm −2 and 24.4 +2.8 −2.0 ×10 14 cm −2 .This indicates a ±15% difference in column density depending on continuum level, and a 20% difference when including the individual errors.Note that the column density of Ne v was based on a Gaussian fit of the troughs of its different energy states, which we further describe in Section 3.2.

Ne v Gaussian Fitting
As seen in Figure 6 (top panel), the Ne v multiplet of the outflow is blended into a singular trough.The involved transitions are Ne v 0 ( = 480.415Å), Ne v* 411 (𝜆 = 481.227, 481.366, 481, 371 Å), and Ne v* 1109 (𝜆 = 482.990, 482.994 Å).To remedy the blending, we modeled the individual energy states of Ne v by fitting Gaussian profiles for each of the expected absorption features, and running a best fit algorithm to best match the data.The free parameters used were the optical depth of the ground state Ne v trough, the width of the trough, and log   .We assumed the AOD scenario, and adjusted the depths of the excited state troughs to match the oscillator strengths of the transition lines, as well as the abundance ratios  (Ne v * )/ (Ne v 0) from the Chianti 9.0.1 atomic database (Dere et al. 1997;Dere et al. 2019).We assumed a temperature of 10,000 K in our Chianti computations.A similar process of finding   via the ratios between the different energy states of Ne v is demonstrated by Miller et al. (2020a), and is especially illustrated in their Figure 3.
We have found that the optical depth  = 0.69 ± 0.11, FWHM = 2360 ± 170 km s −1 and log   = 3.6 ± 0.1 [cm −3 ].Using the modeled troughs, we have calculated the column densities of each energy state of Ne v, as shown in Table 1.Since the value of   is crucial in finding the distance of the outflow from the central source (as described in Section 4.1), we later ran a simulation with the spectral synthesis code Cloudy (version c17.00,Ferland et al. 2017) in order to verify the temperature of the outflow.With the two-phase highionization solution later described in Section 3.3 as our input parameters, the simulation yielded a temperature of  ≈ 27, 000 K. Cal- culating the electron number density with this temperature yielded log   = 4.0 +0.1 −0.1 .As such, we adopted this value of log   for the purpose of our analysis.The total column density of Ne v based on this computation is in agreement with the value based on the  = 10, 000 K assumption.
As an alternate method of modeling the blended trough of the Ne v multiplet, we used the trough of Si xii 499 as a template to create a profile of two blended Gaussians (see Figure 7).We then ran a best fit algorithm to model the absorption of each energy state, leaving the width of the profile as a fixed parameter (see Figure 6 bottom panel).This resulted in an electron number density of log   = 4.3 ± 0.1 [cm −3 ], which is only ∼ 0.3 dex higher than the simple Gaussian fitting shown in Section 3.2.We report the physical properties calculated based on this value of   in Table 5.
While the difference in the electron number density shifts the kinetic luminosity to lower values relative to those shown in Table 3, the kinetic luminosities remain in agreement within error.We thus focus on the results based on the Gaussian model throughout the paper.The parameters are described in further detail in Sections 4.1 and 4.

Photoionization Solution
We used the measured ionic column densities to constrain the values of the hydrogen column density (  ) and ionization parameter (  ), as done in previous works (e.g Xu et al. 2019;Byun et al. 2022a,b,c;Walker et al. 2022).For this purpose, we used a grid of simulated models produced with Cloudy (Ferland et al. 2017) with a range of   and   values as input parameters, modeling the ionic abundances at different   and   .We used the ionic column  The integration range used to calculate the column densities is marked with dotted vertical lines, while the continuum level is indicated by the dashed horizontal line.
densities shown in Table 1 to set upper and lower limits to these parameters, as shown in Figure 9, assuming solar metallicity.We adopted a spectral energy distribution (SED) that would match the V-band flux of Q0254-334 found on NED, the UV continuum flux measured at three separate points, as well as the X-Ray fluxes observed with Chandra at energy ranges from 0.5-7 keV (see Figure 8).Note that there are limitations to this SED due to potential variability between the different observations that were referenced for its construction.Rao et al. (2006) report a V band magnitude of 16 and cite Wright et al. (1982), who in turn discuss observations made with the 3.9 m Anglo-Australian telescope on 28 November, 1978 and on 5 December 1978.Chandra observations of Q0254-334 were made on 2 January, 2000 and 15 February 2000.We have also calculated the   spectral index based on our SED, using the following equation (Tananbaum et al. 1979;Sobolewska et al. 2009): which yielded a result of   = −1.58.This is somewhat higher than the range of   values of LBQS broad absorption line quasars which were reported by Gallagher et al. (2006) (-2.58 to -1.65).
A single phase solution was insufficient to satisfy the constraints from the ionic column densities.To remedy this issue, we formulated a two-phase solution, in which a high-and very high-ionization phase exist co-spatially.We deduced that the two phases would be co-spatial based in the kinematic similarity between the high-ionization troughs and the very high-ionization troughs.Specifically, Figure 6 shows that we get a very good fit for Ne v (a high-ionization line), using the velocity template of Si xii (a very high-ionization line, see Figure 7).We find that the two-phase solution satisfies more of the constraints set by the measured ionic column densities (reduced  2 = 5.1, as opposed to 22.3 for the one-phase solution).To cover the range of possible metallicities, we have also applied models of metallicity  ≈ 4.68 ⊙ (Ballero et al. 2008;Miller et al. 2020b)  values are 16.0 and 0.5, for the one-phase and two-phase solutions respectively.As discussed by Arav et al. (2013), the inability for a one-phase ionization solution to reasonably fit the measurements and limits of   necessitates the adoption of a two-phase solution.This is further demonstrated by comparing the modeled column densities of H i, Na ix, and S xiv from supersolar one-phase and two-phase solutions to the observed ones, as shown in Table 2.Note that the reported measured column density of H i is an upper limit based on Ly .Due to the larger discrepancy between modeled and measured column densities in the solar abundance solutions (e.g.model H i column density upwards of ∼ 20 times larger than measured), they have been excluded from the table .As can be seen in the table, the two-phase solution yields modeled column densities with a maximum 2 difference between modeled and measured column densities, while the one-phase solution yields a 4 − 8 difference.As such, the comparison of modeled column densities favors the two-phase solution.The   and   values found using  2 analysis are shown in Table 3.We compared the   and   values found using the Q0254-334 SED with those found using the SED of the quasar HE0238-1904(hereafter HE0238, Arav et al. 2013), as the latter SED has been adopted for quasar outflow analysis in several past papers (e.g., Miller et al. 2020a;Byun et al. 2022a,b,c;Walker et al. 2022).We report the log   and log   values derived from the HE0238 SED in Table 4. Comparing these values with those found in Table 3 shows that while the one-phase solutions are in agreement within error, the two-phase solutions show a discrepancy in the log   values that range up to ∼ 0.5 dex.

Black Hole Mass Calculation
Black hole masses of AGN are often found using the emission features of Mg ii (Bahk et al. 2019) or C iv (Vestergaard & Peterson 2006;Coatman et al. 2017).However, as the STIS spectrum of Q0254-334 lacked both features, we looked to the O vi emission to compute the mass of the central black hole.We referred to the method described    (Arav et al. 2013) and Q0254-334.The Q0254-334 SED was formed by using the V-band magnitude (the first red dot, Rao et al. 2006), UV continuum flux measured at three different wavelengths (rest wavelengths  = 574, 880, 1097 Å; second, third, and fourth red dots), and the X-ray fluxes reported by Chandra (5th-8th dots).The HE0238 SED was scaled to match the UV continuum flux for the sake of this comparison.We use the Q0254-334 SED for the analysis in this paper.
by Tilton & Shull (2013), measuring the O vi FWHM to find the mass.
Although Tilton & Shull (2013) specify the use of two Gaussians to fit each line of the emission doublet, we opted to fit one Gaussian per line instead, as the lower signal to noise ratio of the STIS spectrum did not warrant the more detailed modeling method.We employed a best fit algorithm adjusting the amplitude of the blue emission line, the ratio between the blue and red lines, and the FWHM of the blue line.The ratio between the blue and red line amplitudes was constrained between 1-1.5, and the widths of the two features were fixed to be equal to each other.For the resulting fit, we found a ratio of 1, normalized amplitude  = 0.23 ± 0.11, and FWHM = 4800 ± 900 km s −1 .This, along with the measured flux of   = 7.7 +1.0 −1.0 × 10 −16 erg s −1 cm −2 Å −1 at rest wavelength 1050Å,  1. Measurements are shown as solid curves, while upper and lower limits are represented with dotted and dashed curves, respectively.The colored bands represent the uncertainties in the constraints.The red circle shows the one-phase solution of   and   , while the black square and star show the high ionization and very high-ionization phase of the two-phase solution respectively.The 1- uncertainties of the solutions are shown as black/red ellipses.resulted in a black hole mass of   = 5.3 +5.5  −2.7 × 10 9  ⊙ , and Eddington luminosity   = 6.6 +6.9 −3.4 × 10 47 erg s −1 .Note that we have limited our Gaussian fit to the red wing of the emission feature, as the blue wing has been contaminated by the absorption outflow (see Fig. 10).While this contamination has contributed significantly to the uncertainty, we were unable to find alternative emission features with which to estimate the black hole mass.

Distance of the Outflow from the Central Source
With the parameters we found as described in Section 3, we could calculate the distance of the outflow from the central source, as well as the kinetic luminosity of the outflow.The distance can be found based on the definition of the ionization parameter   : where   is the emission rate of ionizing photons,  is the outflow distance from the source,   is the hydrogen number density, and  is the speed of light.Solving the equation for  gives us For highly ionized plasma,   ≈ 1.2  (Osterbrock & Ferland 2006), and the values of   and   were found in Section 3. We followed the method of other works (e.g.Miller et al. 2020a;Byun et al. 2022a,c) to find   and integrated over the SED mentioned in Subsection 3.3, limiting our range to energies over 1 Ryd.This yielded the bolometric luminosity   = 2.40 +0.24  −0.24 × 10 47 erg s −1 and   = 9.33 +0.94  −0.94 × 10 56 s −1 .The distance estimates of the outflow calculated with this value are shown in Tables 3  and 5.

Contribution of the Outflow to AGN Feedback
For an outflow to contribute to AGN feedback, its kinetic luminosity must be at least ∼ 0.5% (Hopkins & Elvis 2010) or ∼ 5% (Scannapieco & Oh 2004) of the quasar's Eddington luminosity.Assuming an incomplete spherical shell, the mass flow rate can be calculated as follows: followed by the kinetic luminosity: where Ω is the global covering factor,  = 1.4 is the mean atomic mass per proton,  is outflow velocity, and   is the mass of a proton (Borguet et al. 2012a).We assumed Ω = 0.2, as C iv BALs are found in ∼ 20% of quasars (Hewett & Foltz 2003).We use the Ω associated with C iv BALs, since our high-ionization phase has troughs from ions of very similar ionization potential.For example, in our spectrum, we detect O iv 787.O iv has an ionization potential of 77 eV, which is quite similar to the C iv ionization potential of 64 eV.Assuming supersolar metallicity, this calculation yielded a  * The   of the very high-ionization phase is the   of the high phase times the ratio of the high/very-high ionization parameters.kinetic luminosity of log   = 44.76+0.20 −0.18 [erg s −1 ] for the onephase solution, and log   = 44.98 +0.15  −0.07 [erg s −1 ] for the two-phase solution, leaving a ∼ 0.2 dex difference between the solutions.
The ratio between the kinetic luminosity and Eddington luminosity yields   /  = 0.08 +0.10  −0.05 % for the one-phase solution, and   /  = 0.14 +0.16  −0.07 % for the two-phase solution, which is below the 0.5% threshold.For the sake of completeness, we have also found the ratio between   and the bolometric luminosity   , resulting in   /  = 0.4 +0.3  −0.2 % and 0.7 +0.3 −0.1 %(see Table 3).Based on the ratio between   and   , the outflow would be unable to contribute to AGN feedback.It is important to note that the different assumed metallicity values have significant effects on the physical parameters of the outflow, such as a near order of magnitude difference in kinetic luminosity, leading to values that may be sufficient for AGN feedback contribution (see Table 3).

The Two-Phase Outflow
As mentioned earlier in Section 3.3, the two-phase photoionization solution provides a better fit to the constraints from the measured ionic column densities.While the values of   for the one-phase and two-phase solutions agree with each other within error (see Table 3), there are significant differences to be found in the other parameters, such as distance,   , and   .
Note that the difference in   between the high-and very highionization phases is ∼ 1.5 orders of magnitude, as well as the difference in   .Assuming the two phases are co-spatial, the volume filling factor of the high-ionization phase is as follows (Arav et al. 2013;Miller et al. 2020a):

×
,   ,   (10) resulting in log  V = −3.1 +1.1 −0.9 , which follows our expectations from the high-ionization phase's larger   and smaller   values compared to those of the very high-ionization phase.

Connection to X-Ray Warm Absorbers
The two-phase solution for the outflow of Q0254-334 is comparable to the parameters measured in X-ray warm absorbers.For instance, in their analysis of the Seyfert galaxy NGC 3783, Netzer et al. (2003) found the parameters of the absorbing gas composed of three different components, with the oxygen ionization parameter ranging from log   = −2.4 to −0.6.To effectively compare the   of Q0254-334 to the   values of NGC 3783, we calculated the oxygen ionizing emission rate   as defined below: such that the ratio     =     .The resulting value of the emission rate was   = 3.9 +0.4 −0.4 × 10 54 s −1 , which is 2.4 ± 0.1 orders of magnitude smaller than   .Subtracting 2.4 ± 0.1 from the log   values of the high-and very high-ionization phases leads to log   = −3.2+0.5 −0.5 and −1.6 +0.3 −0.3 respectively.The very high-ionization phase has a   within the range of   values of the NGC 3783 absorbing gas.We note that NGC 3783 is a much lower luminosity AGN than Q0254-334, and that its SED may be different.However, lacking high quality X-ray spectra of  ∼ 1 quasars, it is still illuminating to compare the NGC 3783 X-ray wind with the EUV wind seen in Q0254-334.

Comparison to Other Extreme UV Objects
As the spectrum of Q0254-334 covers observed wavelengths as short as 400Å, we found it appropriate to compare it with other quasars observed in the extreme UV range (hereafter EUV500, Arav et al. 2020).We compiled a list of the physical parameters of 28 EUV500 quasar outflow systems analyzed in previous works (Arav et al. 2020;Xu et al. 2020a,b,c;Miller et al. 2020a,b,c), and added the parameters of Q0254-334 for comparison, with a total of 29 EUV500 outflow systems.Out of the 29 outflow systems, 24, including the outflow discussed in this paper, have measurements of kinetic luminosity and distance from the source.We compared the parameters of the Q0254-334 outflow such as   ,   , , and   , with the other 23 outflow systems.
As seen in Figure 12, no strong correlation has been found between log   and log   , or between log   and log   .4 of the 24 outflows (∼ 16%) are above the threshold of   /  ∼ 5%, while 7 (∼ 29%) are between the 0.5% and 5% thresholds.With regards to   , 5 of the outflow systems (∼ 20%) are above the 5% threshold, while 7 (∼ 29%) are between the 0.5% and 5% thresholds.Note that while the values of log   range between 41-47, log   and log   range between 47.0-47.9and 46.6-47.6 respectively, which is much narrower than the range of log   .It is also indicative of the ability of line-of-sight analysis to identify outflow systems at large ranges of kinetic luminosity, as well as velocity.Figure 13 shows the log   and log   values of the high-and very high-ionization phases of each of the outflow systems.With the exception of the very high-ionization phase of the outflow system of UM425 traveling at −9420 km s −1 , the high-ionization phases tend to have values of log   < 0, while the very high-ionization phases have log   > 0. Note that the very high-ionization phase of the Q0254-334 outflow has a higher log   value relative to the average of the other outflows.We largely attribute this to the detection of S xiv.As can be seen in Figure 9, the very high-ionization phase solution is at the intersection between the Na ix and S xiv constraints.and the parameters were constrained by other ions such as Si xii.It is also notable that as shown in Figure 13, the log   of the very high-ionization phase of the Q0254-334 outflow is higher than the average log   of the other outflows.We suspect that future observed outflows with S xiv would yield comparably high   values.Note that there is an apparent edge in the range of log   and log   values of the outflows.In Figure 13, we have indicated the approximate locations of the hydrogen ionization front (  ≈ 10 23   ), as well as the He ii ionization front (  ≈ 10 22.2   ).The   /  ratio for the He ii ionization front was calculated based on the average log   value at which the He ii to He iii ratio is 1:1 in a series of Cloudy models created with a range of −2.0 < log   < 1.0.We used the aforementioned SED of HE0238 for the models, as this SED was used for the analysis of the majority of the EUV500 outflows in question.Interestingly, the log   vs. log   values of all of the EUV500 outflows fall under the He ii ionization front, which would suggest that they are high ionized BALs (HiBALs) as opposed to low ionized BALs (LoBALs).This is supported by the lack of BALs from low-ionization species.For instance, there is a noticeable lack of absorption where the trough of C ii 687 would be, despite a large oscillator strength of f=0.336 (see Figure 11).
We also examined the ranges of  and  of the outflow systems (see Figure 14).To examine the correlation between distance and velocity, we conducted a weighted least squares linear fit between log  and log ||, taking into account the asymmetry of the reported errors in .We adopted the weight determination method described by Barlow (2003).The weight of each data point   = 1/  was determined by the value of   : .The weighted linear fit yielded a slope of −1.08 and an intercept of 6.44, suggesting a negative correlation.To determine the strength of the correlation, we calculated a modified value of the coefficient of determination  2 that would take into account the weight of each data point.The residual sum of squares    was modified so that: where   is the value of log  according to the linear fit.The total

Figure 1 .Figure 2 .
Figure1.Co-Added STIS spectrum (black)  and FOS spectrum (purple) of Q0254-334.The absorption features of the outflow system ( ≈ −3200 km s −1 ) are marked with red vertical lines, with the Ne v multiplet emphasized with a red arrow.The continuum and emission model is plotted as a blue dashed curve.Note that the continuum flux has risen between 1993 and 2001 in observed wavelengths up to ∼ 2100Å, while at longer wavlengths, the variability is nearly indistinguishable.

Figure 3 .
Figure 3. Absorption troughs of the Q0254-334 outflow plotted in velocity space.The velocity of the outflow at  = 1.89229 is marked with red vertical lines.The integration range used to calculate the column densities is marked with dotted vertical lines, while the continuum level is indicated by the dashed horizontal line.

Figure 4 .Figure 5 .
Figure 4. Absorption troughs of Mg x, Si xii, S xiv, and S iv* in the outflow of Q0254-334.Format and notation are identical to those of Figure 3.

Figure 6 .
Figure 6.Modeling of the Ne v absorption troughs, created by fitting Gaussians (top) and by using Si xii as a template (bottom).The vertical dashed lines represent the range of data used for our fitting.The green curve is the modeled absorption of the resonance state of Ne v.The red curve shows combined absorption of the  = 411 cm − 1 level lines, and the purple curve shows the absorption of the  = 1109 cm −1 level lines.The absorption features from multiple lines of the same excited states have been combined within the figure.The orange curve represents the total combined modeled absorption of the Ne v multiplet.

Figure 7 .Figure 8 .
Figure 7. Fitting of a two-Gaussian profile to the absorption trough of Si xii.The dotted curves show the individual Gaussians in the profile, while the blue curve shows the blended profile of both Gaussians.The dotted vertical lines represent the range of data that was used for fitting the Gaussians.

Figure 9 .
Figure 9. Plots of the Hydrogen column density (  ) vs. ionization parameter (  ), assuming solar (top) and supersolar (bottom) metallicities, with the Q0254-334 SED shown in Figure 8. Constraints on the parameters are based on measured column densities shown in Table1.Measurements are shown as solid curves, while upper and lower limits are represented with dotted and dashed curves, respectively.The colored bands represent the uncertainties in the constraints.The red circle shows the one-phase solution of   and   , while the black square and star show the high ionization and very high-ionization phase of the two-phase solution respectively.The 1- uncertainties of the solutions are shown as black/red ellipses.

Figure 10 .
Figure 10.Gaussian fitting of the O vi emission feature.The dashed vertical lines denote the range of data used for the Gaussian fit.The green and red curves are the modeled blue and red emission features respectively, and the orange curve represents the combined modeled emission.
of the upper and lower errors of log , while   =  +  −  −  2

Figure 12 .
Figure 12.Distribution of log   vs. log   (left) and log   (right) of EUV500 outflows.The dashed and solid lines on the left (right) indicate the   /  (   /  ) thresholds of 0.5% and 5% respectively.The plus-sign symbol denotes the outflow of Q0254-334 as reported in this paper, while the other symbols denote the parameters of other EUV outflows reported byArav et al. (2020) andMiller et al. (2020c).The color map corresponds to the velocities of the outflow systems.

Figure 13 .
Figure 13.Distribution of log   vs. log   of EUV500 outflows.Symbols are coded as they are in Figure 12.Symbols with dotted outlines denote high-ionization phases, while symbols with solid outlines denote very highionization phases.The color map corresponds to the velocities of the outflow systems.The dark blue and cyan curves show the H i and He ii ionization fronts respectively.

Table 1 .
Q0254-334 outflow column densities from STIS observations.The numbers next to the Ne v* excited states denote the energies in cm −1 .The values are in units of 10 14 cm −2 .
, which are shown in the lower panel of Figure 9.The results are favorable towards the super-solar metallicity solution, of which the reduced  2 (d) S iv

Table 2 .
The measured and modeled column densities of H i, Na ix, and S xiv of the Q0254-334 outflow.The second and third columns denote the supersolar one-phase solution and the  difference between modeled and measured values; the fourth and fifth columns show the same for the twophase solution.The values are in units of 10 14 cm −2 .

Table 3 .
Physical Properties of the Q0254-334 Outflow.The high and very high ionization phases for the two-phase solution are assumed to be co-spatial.
The log   value is ∼ 0.3 dex higher than what it would have been if S xiv were not detected, Region in which C ii 687 absorption is expected to be found.