The Relation between Galaxy ISM and Circumgalactic O vi Gas Kinematics Derived from Observations and ΛCDM Simulations

The HST/COS quasar spectra have a resolution of R ∼ 20,000 and cover the O vi λλ1031, 1037 doublet for the targeted galaxies. Details of the HST/COS observations are presented in Kacprzak et al. (2015). The data were reduced using the CALCOS software. Individual grating integrations were aligned and co-added using the IDL code "coadd_x1d" developed by Danforth et al. (2010).⁶ Since the COS FUV spectra are oversampled (six pixels per resolution element), we binned the data by three pixels to increase the signal-to-noise ratio, and all of our analysis was performed on the binned spectra. Continuum normalization was performed by fitting the absorption-free regions with smooth low-order polynomials.

We adopted the fitted rest-frame EWs and column densities from Kacprzak et al. (2015). Non-Gaussian line-spread functions (LSFs) were adopted and obtained by interpolating the LSF tables (Kriss 2011) at the observed central wavelength for each absorption line and were convolved with the fitted model Voigt profile VPFIT (Carswell & Webb 2014). In all cases, a minimum number of components was used to obtain a satisfactory fit with reduced χ² ∼ 1. The O vi λ1031 model profiles were used to compute the EWs, and the 1σ errors were computed using the error spectrum. Both the EWs and column densities are listed in Table 1.

Table 1.  Absorption and Host Galaxy Properties

Quasar	zabs	zgala	HST	mHST	log(Mh)	Rvir	D	i	Φ	EWr	log N(O vi)
Field			Filter	(AB)	(M⊙)	(kpc)	(kpc)	(deg)	(deg)	(Å)
Quasar Sight lines Located along the Galaxy's Major Axis (Φ < 25°)
J035128.54−142908.7	0.356825	0.356992	F702W	20.7	${12.0}_{-0.2}^{+0.3}$	${191}_{-26}^{+48}$	72.3 ± 0.4	${28.5}_{-12.5}^{+19.8}$	${4.9}_{-40.2}^{+33.0}$	0.396 ± 0.013	14.76 ± 0.17
J091440.39+282330.6	0.244098	0.244312	F814W	19.6	${11.9}_{-0.2}^{+0.3}$	${171}_{-24}^{+49}$	105.9 ± 0.1	${39.0}_{-0.2}^{+0.4}$	${18.2}_{-1.0}^{+1.1}$	0.333 ± 0.028	14.65 ± 0.07
J094331.61+053131.4	0.353286	0.353052	F814W	21.2	${11.7}_{-0.2}^{+0.4}$	${147}_{-22}^{+54}$	96.5 ± 0.3	${44.4}_{-1.2}^{+1.1}$	${8.2}_{-5.0}^{+3.0}$	0.220 ± 0.024	14.66 ± 0.07
J095000.73+483129.3	0.211757	0.211866	F814W	18.0	${12.4}_{-0.2}^{+0.2}$	${247}_{-29}^{+36}$	93.6 ± 0.2	${47.7}_{-0.1}^{+0.1}$	${16.6}_{-0.1}^{+0.1}$	0.211 ± 0.019	14.32 ± 0.04
J104116.16+061016.9	0.441630	0.442173	F702W	20.9	${12.0}_{-0.2}^{+0.3}$	${193}_{-25}^{+42}$	56.2 ± 0.3	${49.8}_{-5.2}^{+7.4}$	${4.3}_{-1.0}^{+0.9}$	0.368 ± 0.023	14.64 ± 0.18
J113910.79−135043.6	0.204297	0.204194	F702W	20.0	${11.7}_{-0.2}^{+0.4}$	${146}_{-22}^{+52}$	93.2 ± 0.3	${81.6}_{-0.5}^{+0.4}$	${5.8}_{-0.5}^{+0.4}$	0.231 ± 0.009	14.40 ± 0.28
J132222.46+464546.1	0.214320	0.214431	F814W	18.6	${12.1}_{-0.2}^{+0.3}$	${205}_{-26}^{+44}$	38.6 ± 0.2	${57.9}_{-0.2}^{+0.1}$	${13.9}_{-0.2}^{+0.2}$	0.354 ± 0.024	14.62 ± 0.12
J134251.60−005345.3	0.227196	0.227042	F814W	18.2	${12.4}_{-0.2}^{+0.2}$	${252}_{-29}^{+36}$	35.3 ± 0.2	${10.1}_{-10.1}^{+0.6}$	${13.2}_{-0.4}^{+0.5}$	0.373 ± 0.023	14.58 ± 0.11
J213135.26−120704.8	0.430164	0.430200	F702W	20.7	${12.0}_{-0.2}^{+0.3}$	${200}_{-25}^{+42}$	48.4 ± 0.2	${48.3}_{-3.7}^{+3.5}$	${14.9}_{-4.9}^{+6.0}$	0.385 ± 0.013	14.60 ± 0.05
J225357.74+160853.6	0.390705	0.390013	F702W	20.6	${12.2}_{-0.2}^{+0.2}$	${217}_{-28}^{+45}$	276.3 ± 0.2	${76.1}_{-1.2}^{+1.1}$	${24.2}_{-1.2}^{+1.2}$	0.173 ± 0.030	14.29 ± 0.04
Quasar Sight lines Located along the Galaxy's Minor Axis (${\rm{\Phi }}\gt 33^\circ$)
J012528.84−000555.9	0.399090	0.398525	F702W	19.7	${12.5}_{-0.2}^{+0.2}$	${285}_{-32}^{+37}$	163.0 ± 0.1	${63.2}_{-2.6}^{+1.7}$	${73.4}_{-4.7}^{+4.6}$	0.817 ± 0.023	15.16 ± 0.04
J045608.92−215909.4	0.381514	0.381511	F702W	20.7	${12.0}_{-0.2}^{+0.3}$	${192}_{-26}^{+48}$	103.4 ± 0.3	${57.1}_{-2.4}^{+19.9}$	${63.8}_{-2.7}^{+4.3}$	0.219 ± 0.013	14.34 ± 0.13
J094331.61+053131.4	0.548769	0.548494	F814W	21.0	${12.0}_{-0.2}^{+0.3}$	${191}_{-25}^{+43}$	150.9 ± 0.6	${58.8}_{-1.1}^{+0.6}$	${67.2}_{-1.0}^{+0.9}$	0.275 ± 0.050	14.51 ± 0.07
J100902.07+071343.9	0.227851	0.227855	F625W	20.1	${11.8}_{-0.2}^{+0.4}$	${155}_{-23}^{+51}$	64.0 ± 0.8	${66.3}_{-0.9}^{+0.6}$	${89.6}_{-1.3}^{+1.3}$	0.576 ± 0.021	15.14 ± 0.10
J113910.79−135043.6	0.212237	0.212259	F702W	20.0	${11.7}_{-0.2}^{+0.4}$	${150}_{-22}^{+52}$	174.8 ± 0.1	${85.0}_{-0.6}^{+0.1}$	${80.4}_{-0.5}^{+0.4}$	0.137 ± 0.009	14.12 ± 0.12
J113910.79−135043.6	0.319167	0.319255	F702W	20.6	${11.9}_{-0.2}^{+0.3}$	${170}_{-24}^{+51}$	73.3 ± 0.4	${83.4}_{-1.1}^{+1.4}$	${39.1}_{-1.7}^{+1.9}$	0.255 ± 0.012	14.41 ± 0.09
J124154.02+572107.3	0.205538	0.205267	F814W	19.9	${11.6}_{-0.2}^{+0.4}$	${140}_{-21}^{+52}$	21.1 ± 0.1	${56.4}_{-0.5}^{+0.3}$	${77.6}_{-0.4}^{+0.3}$	0.519 ± 0.018	14.89 ± 0.13
J155504.39+362847.9	0.189033	0.189201	F814W	18.5	${12.1}_{-0.2}^{+0.3}$	${194}_{-25}^{+45}$	33.4 ± 0.1	${51.8}_{-0.7}^{+0.7}$	${47.0}_{-0.8}^{+0.3}$	0.385 ± 0.033	14.74 ± 0.17
J225357.74+160853.6	0.153821	0.153718	F702W	19.3	${11.6}_{-0.2}^{+0.5}$	${130}_{-20}^{+53}$	31.8 ± 0.2	${59.6}_{-1.7}^{+0.9}$	${33.3}_{-1.9}^{+2.7}$	0.263 ± 0.056	14.59 ± 0.06
J225357.74+160853.6	0.352708	0.352787	F702W	20.3	${11.9}_{-0.2}^{+0.3}$	${180}_{-25}^{+50}$	203.2 ± 0.5	${36.7}_{-4.6}^{+6.9}$	${88.7}_{-4.8}^{+4.6}$	0.381 ± 0.036	14.70 ± 0.15

Note.

^aKeck ESI redshifts derived from this work.

Download table as:  ASCIITypeset image

All quasar/galaxy fields have been imaged with HST using the Advanced Camera for Surveys (ACS), WFC3, or WFPC2. Details of the observations are found in Kacprzak et al. (2015), and the filters used are found in Table 1. ACS and WFC3 data were reduced using the DrizzlePac software (Gonzaga et al. 2012). When enough frames were present, cosmic rays were removed during the multidrizzle process; otherwise, L.A.Cosmic was used (van Dokkum 2001). WFPC2 data were reduced using the WFPC2 Associations Science Products Pipeline (see Kacprzak et al. 2011b).

Galaxy photometry was adopted from Kacprzak et al. (2015), who used the Source Extractor software (SExtractor; Bertin & Arnouts 1996) with a detection criterion of 1.5σ above background. The m_HST magnitudes in each filter are quoted in the AB system and are listed in Table 1.

We adopt calculated halo masses and virial radii from Ng et al. (2019), who applied halo abundance matching methods in the Bolshoi N-body cosmological simulation (Klypin et al. 2011); see Churchill et al. (2013a, 2013b) for further details.

The galaxy morphological parameters and orientations are adopted from Kacprzak et al. (2015). In summary, morphological parameters were quantified by fitting a two-component disk+bulge model using GIM2D (Simard et al. 2002), where the disk component has an exponential profile while the bulge has a Sérsic profile (Sérsic 1968) with 0.2 ≤ n ≤ 4.0. The galaxy properties are listed in Table 1. We use the standard convention of the azimuthal angle Φ = 0° to be along the galaxy-projected major axis and Φ = 90° to be along the galaxy-projected minor axis.

The galaxy spectra were obtained using the Keck Echelle Spectrograph and Imager (ESI; Sheinis et al. 2002). The ESI slit position angle was selected to be near the optical major axis of each galaxy in order to accurately measure the galaxy rotation curves. Details of the ESI/Keck observations are presented in Table 2. The ESI slit is 20'' in length and set to 1'' wide. We binned by two in the spatial directions, resulting in pixel scales of 027–034 over the echelle orders of interest. Binning by two in the spectral direction results in a sampling rate of 22 km s⁻¹ pixel⁻¹ (FWHM ∼ 90 km s⁻¹). ESI has a wavelength coverage of 4000–10000 Å, which covers multiple emission lines such as the [O ii] doublet, Hβ, the [O iii] doublet, Hα, and the [N ii] doublet.

Table 2.  ESI Observations

Quasar	${z}_{\mathrm{gal}}^{{\prime} }$	${z}_{\mathrm{gal}}^{{\prime} }$	Observation	Slit PA	Exp
Field		Refa	Date	(deg)	(s)
J012528.84−000555.9	0.3985	3	2014 Dec 13	15	1800
J035128.54−142908.7	0.3567	1	2014 Dec 13	110	2550
J045608.92−215909.4	0.3818	1	2014 Dec 13	110	1200
J091440.39+282330.6	0.2443	2	2016 Jan 15	23	1500
J094331.61+053131.4	0.3530	2	2016 Jan 15	38	4500
J094331.61+053131.4	0.5480	2	2016 Jan 15	−218	4500
J095000.73+483129.3	0.2119	2	2016 Jan 15	13	1000
J100902.07+071343.9	0.2278	2	2016 Jan 15	115	1200
J104116.16+061016.9	0.4432	5	2014 Apr 25	−110	3300
J113910.79−135043.6	0.2044	1	2016 Jan 15	160	1650
J113910.79−135043.6	0.2123	1	2016 Jan 15	111	2400
J113910.79−135043.6	0.3191	1	2016 Jun 6	94	1800
J124154.02+572107.3	0.2053	2	2016 Jun 6	121	1200
J132222.46+464546.1	0.2142	2	2016 Jun 6	0	1500
J134251.60−005345.3	0.2270	2	2016 Jun 6	−24	1800
J155504.39+362847.9	0.1893	2	2016 Jun 6	130	1800
J213135.26−120704.8	0.4300	6	2015 Jul 16	55	6000
J225357.74+160853.6	0.1530	1	2015 Jul 16	−213	3000
J225357.74+160853.6	0.3526	1	2016 Jun 6	−185	3300
J225357.74+160853.6	0.3900	1	2016 Jun 6	30	1200

Note.

^aOriginal galaxy redshift (${z}_{\mathrm{gal}}^{{\prime} }$) source: (1) Chen et al. (2001), (2) Werk et al. (2012), (3) Muzahid et al. (2015), (4) Kacprzak et al. (2010), (5) Steidel et al. (2002), (6) Guillemin & Bergeron (1997).

Download table as:  ASCIITypeset image

All ESI data were reduced using IRAF. Galaxy spectra are both vacuum and heliocentric velocity corrected to provide a direct comparison with the absorption-line spectra. The derived wavelength solution was verified against a catalog of known sky lines that resulted in an rms difference of ∼0.03 Å (∼2 km s⁻¹).

The galaxy rotation curve extraction was performed following the methods of Kacprzak et al. (2010, 2011a; see also Vogt et al. 1996; Steidel et al. 2002). We extracted multiple spectra along the spatial direction of a galaxy using 3-pixel-wide apertures (corresponding to approximately one resolution element of 081–102 for ESI) at 1-pixel spatial increments along the slit. To obtain accurate wavelength calibrations, we extract spectra of the arc lines at the same spatial pixels as the extracted galaxy spectra. Fitted arc lamp exposures provided a dispersion solution with an rms of ∼0.035 Å (∼2 km s⁻¹). The Gaussian fitting algorithm (FITTER; see Churchill et al. 2000) was used to compute best-fit emission- and absorption-line centers and widths to derive galaxy redshifts and kinematics. Galaxy redshifts were computed at the velocity centroid of the line, accounting for emission-line resolved kinematics and/or luminosity asymmetries. The galaxy redshifts are listed in Table 1; their accuracy ranges from 3 to 20 km s⁻¹. The 20 rotation curves are presented in Appendix A for the 10 quasar sightline along the galaxy's major axis (Figures 12–16) and in Appendix B for the 10 quasar sightline along the galaxy's minor axis (Figures 17–21).

We first explore whether the O vi CGM gas is gravitationally bound to their host galaxy dark matter halos. In Figure 1, we show the velocity difference between the median optical depth distribution of the O vi λ1031 absorption line and the galaxy systemic velocity as a function of the host galaxy halo mass for all azimuthal angles. The error bars show the full velocity range of the absorption, which is defined as where the Voigt profile fitted absorption models return to 1% from the continuum level. The Voigt profile models are preferred to define the velocity ranges since some O vi absorption systems are blended with other ions in the spectra (see Nielsen et al. 2017) and the data tend to be quite noisy.

Zoom In Zoom Out Reset image size

The rest-frame velocity differences between galaxies and their associated O vi absorption have a mean offset of dv = 9.2 ± 58.8 km s⁻¹, with standard error of the mean of 13.5 km s⁻¹. This implies that most of the gas resides near the galaxy systemic velocity regardless of its orientation with respect to the host galaxy. Also included in the figure are curves indicating the escape velocity for a given halo mass at an impact parameter of D = 200 kpc (inner curve) and 100 kpc (outer curve) at the median redshift of z = 0.3. Note that little to no absorption resides outside of these curves, indicating that the O vi gas is bound to their dark matter halos. These results are consistent with previous findings showing bound O vi gas (e.g., Tumlinson et al. 2011; Stocke et al. 2013; Mathes et al. 2014).

Given the observed O vi azimuthal angle bimodality (Kacprzak et al. 2015), our sample can be easily split into two azimuthal angle bins considered as major- and minor-axis samples. Here we discuss a subset of 10 systems where the O vi absorption is detected within 25° of the galaxy major axis. This major-axis azimuthal cut was selected to mimic the Mg ii major-axis sample of Ho et al. (2017).

We aim to determine whether major-axis O vi displays similar corotation kinematic signatures to those commonly seen for Mg ii absorption (e.g., Steidel et al. 2002; Kacprzak et al. 2010; Ho et al. 2017). In Figure 2, we present the data used in this analysis for two example fields J0351 and J0914. The figures for the targeted galaxies where the quasar sightline aligns with their major axis are located in Appendix A (Figures 12–16). Figure 2 shows the O vi host galaxies and the quasars in the HST images along with the ESI slit position placed over each galaxy. The figure further shows the Hα-derived galaxy rotation curve, obtained from the ESI spectra, and the HST/COS O vi absorption profile. All velocities are shown with respect to the galaxy systemic velocity. Note that the rotation speeds are low (∼50 km s⁻¹), which is expected for these moderately inclined spiral galaxies. For the galaxy in J0351, the O vi absorption profile covers the entire kinematic range of the galaxy rotation curve. As for the galaxy in J0914, the O vi absorption resides mostly to one side of the galaxy systemic velocity as previously seen for Mg ii systems. We now explore whether corotating/lagging disk models can explain the observed CGM kinematics.

Zoom In Zoom Out Reset image size

Similar to previous works, we apply the simple monolithic halo model from Steidel et al. (2002) to determine whether an extended disk-like rotating gas disk (as commonly seen for Mg ii) is able to reproduce the observed O vi absorption velocity spread given the galaxy's rotation speed and relative orientation with respect to the quasar sightline. In summary, model line-of-sight velocities (v_los) are a function of the measurable quantities of impact parameter (D), galaxy inclination angle (i), galaxy–quasar position angle (Φ), and the maximum projected galaxy rotation velocity (v_max) such that

$$\begin{eqnarray}\begin{array}{rcl}{v}_{\mathrm{los}} & = & \displaystyle \frac{-{v}_{\max }}{\sqrt{1+{\left(\tfrac{y}{p}\right)}^{2}}}\,\exp \left\{-\displaystyle \frac{\left|y-{y}_{\circ }\right|}{{h}_{v}\tan i}\right\}\,\,{\rm{where}},\\ {y}_{\circ } & = & \displaystyle \frac{D\sin {\rm{\Phi }}}{\cos i}\ \mathrm{and}\ p\,=\,D\cos {\rm{\Phi }},\end{array}\end{eqnarray} \tag{ 1 }$$

where h_v is a free parameter representing the scale height for the velocity lag of the CGM. Here we assume a thick disk (h_v = 1000 kpc), which represents the maximum disk/CGM rotation scenario. Assuming a maximum disk rotation model is reasonable given that we do not know how/if the velocity changes with impact parameter and there is little to no velocity gradient along the corotating gaseous structures within the simulations (e.g., Stewart et al. 2011, 2013, 2017).

The parameter y is the projected line-of-sight position above the disk plane, and y_o is the position at the projected disk midplane. The distance along the sightline relative to the point where it intersects the projection of the disk midplane is D_los = (y − y_o)/sini. Thus, D_los = 0 kpc is where the model line-of-sight intersects the projected midplane of the galaxy. Please see Figure 6 from Steidel et al. (2002) for a visual representation of the model.

Shown in the bottom panels of Figure 2 are the line-of-sight velocities through the halo derived for the geometry of both galaxy–quasar pairs for CGM gas rotating at a maximum velocity set by the rotation curves (solid curves). The dashed curves indicate model velocities derived from uncertainties in i and Φ. In most cases, error values are small such that the dashed curves lie near/on the solid curves. The z = 0.3570 galaxy in the J0351 field has most of the O vi absorption blueward of the galaxy systemic velocity, which agrees with the direction of rotation of the galaxy and thus the model halo. However, given the galaxy's moderate inclination, the model is unable to account for the large velocity spread measured for the absorption profile. Similarly, the z = 0.2443 galaxy in the J0914 field also has the majority of the O vi blueward of the galaxy systemic velocity. In this case, the galaxy is consistent with being counterrotating with respect to the O vi, with this model again failing to reproduce the observed velocity spread.

Figure 3 shows the O vi λλ1031, 1037 absorption profiles along with the Voigt profile fits for the 10 galaxies that have quasar sightlines passing within 25° of the host galaxy's projected major axis. The absorption profiles are plotted relative to the host galaxy systemic velocities. Note that the bulk of the gas resides near the galaxy systemic velocity with a relatively large velocity spread. Below the profiles are the modeled corotating line-of-sight velocities. It is immediately clear that the absorption profiles have a much higher velocity range than that of the predicted model line-of-sight velocities. Furthermore, only four systems (J0351, J0943, J1041, and J2131) have models that are rotating in the direction of the bulk of the O vi but still fall short of predicting the observed velocities. The z = 0.35 system in the J0943 field is the only system where the observed O vi gas could be explained by disk rotation and/or accretion. Five systems are consistent with counterrotating O vi absorption relative to their host galaxies (J0914, J0950, J1322, J1342, and J2251). These results are in stark contrast to what has been found for Mg ii galaxy–absorption pairs.

Zoom In Zoom Out Reset image size

Given that the quasar sightlines are within 25° of the galaxy major axis, we created the top panel of Figure 4, which shows the rotation velocities as a function of the projected distance between the galaxies and their quasar sightlines. The rotation curves for each galaxy are orientated such that the quasar sightline are located along the positive velocity arm of the rotation curves. The O vi is shown with respect to the galaxy systemic velocity and has the same color as plotted for the rotation curve of their host galaxy. The error bars indicate the full extent of the absorption, while the shaded region shows the actual absorption profile in velocity space. This allows the reader to see where the bulk of the optical depth is in relation to the galaxy rotation. The white tick mark indicates the optical-depth-weighted median of the O vi absorption profile.

Zoom In Zoom Out Reset image size

It can clearly be seen in the top panel of Figure 4 that the majority of O vi is inconsistent with corotation and gas accretion models. Most of the gas resides near the galaxy systemic velocity, and there is no preference toward the direction of galaxy rotation. It is still plausible that some of the gas could be accreting/corotating; however, the signature is not strong or is masked by the other kinematics ongoing within the halo. A mean stacked spectrum of all 10 absorption systems is also shown in the top panel, where the O vi absorption almost symmetrically spans the galaxy systemic velocity where the optical-depth-weighted median of the stacked spectrum is at 2.5 km s⁻¹ relative to the galaxy systemic velocity. Furthermore, the O vi profile encompasses the entire rotational dynamics of the galaxies. This is more clearly shown in the middle panel of Figure 4, where all the velocities are normalized to the maximum line-of-sight rotation velocity for each rotation curve and their associated absorption. Almost all of the systems span the entire dynamical range of the galaxy rotation and more. Again, a mean stacked spectrum is also shown, and the O vi spans more than twice the full dynamic range of the rotation curves. Recall, though, that the O vi gas is still gravitationally bound to their halos. The bottom panel in Figure 4 is similar to the middle panel, except now shown as a function of viral radius derived for each galaxy. This clearly shows that 9/10 systems are well within the viral radius.

These results indicate that a rotating disk and/or radial infall does not provide a plausible explanation for the total observed O vi kinematics. Thus, this clearly indicates that if there exists a kinematic connection between highly ionized gas and its galaxies, then it is very low and/or masked by other kinematic sources such as diffuse gas found within the halo. Given that the quasar sightlines are within <25° of the galaxy major axes, ongoing outflows would not likely contribute to the absorption kinematics seen here. However, it is possible that recycled gas could could be dominating the observed kinematics.

Here we discuss a subset of 10 systems where the O vi absorption is detected at >33° from the galaxy major axis (within 57° of the galaxy minor axis). This angle was selected given that O vi outflowing gas could likely occur within half-opening angles as small as 30° or even larger to 50° (Kacprzak et al. 2015). Figures 17–21 show the HST images along with the galaxy rotation curves and their corresponding O vi absorption. Inspection of these figures shows that the absorption spans the entire galaxy systemic velocity and encompasses the full galaxy rotation velocity range in 4/10 cases, while 6/10 systems have most of the O vi absorption offset to one side of the galaxy systemic velocity.

Previous studies have shown that Mg ii galaxy–absorber pairs with disk-like rotation can be found for quasar sightlines with intermediate to high Φ values (e.g., Kacprzak et al. 2010). We explore the disk-like rotation model from Equation (1) for our minor-axis sample in Figure 5. In Figure 5, the fitted O vi doublet is shown along with the modeled line-of-sight velocities through the halo derived for each galaxy using the maximum velocity set by the rotation curves (solid curves). We find three systems (J1241, J0943, and J1555) where the model can account for all the observed absorption velocities. However, 7/10 have model kinematics consistent with counterrotation with respect to the bulk of the O vi absorption. What these models demonstrate is that there is not an overall consistency between the relative O vi-galaxy kinematics. Only 4/20 from the total sample of major- and minor-axis galaxies have relative velocities expected of disk-like rotation/gas accretion. This is in stark contrast to the commonly observed corotation found for Mg ii absorption.

Zoom In Zoom Out Reset image size

Given the relative quasar–galaxy geometry, it could be expected that outflows might be commonly observed along the galaxy minor axis. Furthermore, this outflowing gas is likely traced by warm O vi absorption. To test this, we apply a simple conical model for outflowing gas from Gauthier & Chen (2012). Here we summarize the model, but see Gauthier & Chen (2012) for details and their Figure 1 for an illustration of the model.

Their collimated outflow model is characterized by an expanding cone originating from the galaxy center along the polar axis with a total angular span of 2θ₀. As with the disk rotation model, i is the inclination of the galaxy, while Φ is the angle between the projected major axis of the disk and the quasar sightline that is at an impact parameter D. These measured quantities are found in Table 1. The quasar line of sight intercepts the outer edges of the outflow cone at a height z from z₁ to z₂, which is determined by the cone opening angle θ₀. The position angles, ϕ_[1,2], of the projected outflow cross section at z_[1,2] are constrained by

$$\begin{eqnarray}&&\tan {\phi }_{[\mathrm{1,2}]}=\displaystyle \frac{D\,\sin \alpha -{z}_{[\mathrm{1,2}]}\,\sin i}{D\,\cos \alpha },\end{eqnarray} \tag{ 2 }$$

and the relation between z_[1,2] and the opening angle θ₀ is

$$\begin{eqnarray}&&{z}_{[\mathrm{1,2}]}\,\tan {\theta }_{0}=D\sqrt{1+{\sin }^{2}{\phi }_{[\mathrm{1,2}]}{\tan }^{2}i}\left(\displaystyle \frac{\cos {\rm{\Phi }}}{\cos {\phi }_{[\mathrm{1,2}]}}\right).\end{eqnarray} \tag{ 3 }$$

Equations (2)–(3) can be used to calculate the corresponding θ at any given point along the quasar sightline at height z where z₁ ≤ z ≤ z₂.

The outflow speed, v, of a gas cloud moving outward at a height z corresponds to the line-of-sight velocity v_los such that

$$\begin{eqnarray}&&v=\displaystyle \frac{{v}_{\mathrm{los}}}{\cos j},\,\,{\rm{where}}\ \ j={\sin }^{-1}\left(\displaystyle \frac{D}{z}\cos \theta \right).\end{eqnarray} \tag{ 4 }$$

The line-of-sight velocities are defined by the redmost and bluemost velocity edges of the absorption profile relative to the galaxy systemic velocity. The gas producing the observed absorption is assumed to be distributed symmetrically around the polar axis of the cone, and the absorption at z₁ and z₂ probes regions close to the front and back side of the outflow, respectively. If asymmetry arises owing to inhomogeneities of gas with the outflows, then the computed velocity gradients represent a lower limit to the outflow velocity field.

We apply the above model since our observational data of the galaxy–quasar geometry provide constraints on θ₀ and the absorption profiles constrain the plausible outflow velocities. From these data, we can identify whether or not reasonable opening angles and outflow velocities are able to replicate the observations. If so, then outflows are a plausible explanation for the observed kinematics, and if not, then outflows may not be the likely source driving the O vi gas kinematics seen along the galaxy minor axis.

Figure 5 shows the outflow models for each quasar–galaxy pair for their relative orientation and absorption gas–galaxy kinematics. The top right panel for each system shows at what height the quasar sightline enters the outflow cone (blue dashed and solid lines) and what height the sightline exits the cone (red dashed and solid lines) as a function of outflow opening angle. The figure shows scenarios where the opening angle is not well constrained, as seen for J1139_0.2123, since the galaxy is nearly edge-on (i = 85°) and the quasar sightline is almost directly along the minor axis (within 96). The other scenario shown is for galaxies where the quasar sightline is not directly along the galaxy minor axis and the outflow opening angle has to be sufficiently large enough before it intercepts the sightline. This can be seen for J1139_0.3193 (Φ = 39°) and for J2253_0.1537 (Φ = 33°), where the opening has to be at least 50° before the sightline intercepts the cone.

In all cases, the opening angle on both sides of the cone can be large enough that the quasar sightline no longer intercepts the cone, which is why z asymptotes to large values. We use the far side of the cone (red) as an upper limit on the outflow half-opening angle. From geometric arguments only, the model constrains the half-opening angles to range from 0°–50° as the smallest possible angle to 26°–83° at its largest. The half-opening angle model results are presented in Table 3. These are consistent with expected/modeled values found for cooler gas tracers, which range between 10° and 70° (Walter et al. 2002; Gauthier & Chen 2012; Kacprzak et al. 2012; Martin et al. 2012; Bordoloi et al. 2014). These are also consistent with those derived by Kacprzak et al. (2015), who examined the azimuthal angle dependence of the gas covering fraction and concluded that the O vi outflowing gas could occur within a half-opening angle as small as 30° or even larger at 50°.

Table 3.  Model Outflow Half-opening Angles and Velocities

Quasar	${z}_{\mathrm{abs}}$	z(i)a	vred(i)b	vblue(i)c	θ (deg)d	θ (deg)e	θ (deg)f
Field		(kpc)	( km s−1)	( km s−1)	(Geometric)	(Velocity Limited)	(Acceleration Only)
J012528.84−000555.9	0.398525	174	991	406	15–62	15–23 and 40.1–62	15–19
J045608.92−215909.4	0.381511	118	266	279	22–57	22–34 and 49.0–57	⋯
J094331.61+053131.4	0.548494	169	369	168	19–58	19–34 and 42.4–58	19–23
J100902.07+071343.9	0.227855	70	540	450	0–66	0–14 and 33.8–66	0–3
J113910.79−135043.6	0.212259	172	>1000	>1000	10–81	15–81	⋯
J113910.79−135043.6	0.319255	47	>1000	996	50–83	53–83	65–83
J124154.02+572107.3	0.205267	26	428	223	11–55	11–29 and 42–55	11–16
J155504.39+362847.9	0.189201	43	277	158	33–50	33–50	33–36
J225357.74+160853.6	0.153718	31	162	74	46–60	46–60	46–50
J225357.74+160853.6	0.352787	342	192	184	1–26	1–26	⋯

Notes.

^aThe height above the disk. ^bThe velocity of the red side of the cone at the lowest value of the opening angle. ^cThe velocity of the blue side of the cone at the lowest value of the opening angle. ^dThe half-opening angle constrained by geometric arguments only. ^eThe half-opening angle constrained by geometric arguments and for velocities less than 1000 km s⁻¹. ^fHalf-opening angles where accelerated outflows exist.

Download table as:  ASCIITypeset image

If realistic outflow velocities can reproduce the observed absorption, it would be a key step for understanding whether outflows can explain the observed O vi gas–galaxy kinematics. The bottom panels in Figure 5 show the model outflow velocities at the edges of the cones required to reproduce the entire velocity spread of the observed O vi absorption profile with respect to the galaxy systemic velocity. The blue line corresponds to the outflow velocities where the quasar sightline enters the outflow cone, while the red line corresponds to the outflow velocities where the quasar sightline exits the outflow cone. Each galaxy–absorber pair has a large range of modeled velocities as a function of opening angle required to reproduce the observed line-of-sight velocities. Some of these velocities far exceed 1000 km s⁻¹ as the dot product of the outflow velocity vector and the line-of-sight velocity vector approaches zero. Here we assume that the outflow velocities at large distances above the galaxy disk likely do not exceed 1000 km s⁻¹ for these systems. This assumption provides additional constraints on the acceptable outflow geometry indicated by the solid line, and those values are listed in Table 3. While the range in opening angles is more limited, the viable half-opening angles are still consistent with previous works.

The ranges of the outflow velocities required to reproduce the observed O vi kinematics shown in Figure 5 appear reasonable and are within a few hundred kilometers per second—typical of expected outflow velocities (e.g., Martin & Bouché 2009; Weiner et al. 2009; Bouché et al. 2012; Gauthier & Chen 2012; Martin et al. 2012; Rubin et al. 2014; Schroetter et al. 2016). We emphasize, however, that a modeled active accelerating outflow occurs when the line-of-sight velocity increases from where the quasar sightline enters the conical outflow to where the quasar sightline exits the conical outflow (since z₁ < z₂). The outflow velocities (v) shown in Figure 5 would be consistent with an active outflowing model when the blue line is below the red line. In the opposite case, where the red line is below the blue line, the outflow is decelerating in order to reproduce the observed kinematics.

We find that 7/10 galaxies exhibit outflowing gas with an accelerated flow. Note, however, that acceleration only occurs for very small opening angles typically within 20° and only over a range of 10° (with the exception of J1139_0.3193). These values are also listed in Table 3. Thus, if active outflows are occurring, they only occur within a very small opening angle conical outflow.

For the majority of the opening angle range, the outflow velocities required to reproduce the observations would be decelerating as the gas moves farther away from the host galaxy. With the assumed velocity cut of 1000 km s⁻¹, there still remains a large range of opening angles that are valid (see Table 3). This would imply that either active outflows exist, and at these large heights above the disk the gas is rapidly decelerating, or the absorption is a result of previously ejected gas that is potentially falling back onto the galaxy.

A caveat of these models is that we have assumed that all of the gas seen in absorption is a result of the outflow. If only some fraction of the gas is associated with outflows, then the model velocities, and where acceleration and deceleration occur, would be different and likely are expressed as upper limits. However, we do not have any evidence to counter this assumption. Thus, we find that accelerating outflow gas can only occur over a very small range of opening angles, and most of the time the gas is found to be decelerating.

We analyzed ΛCDM cosmological simulations created using the Eulerian gasdynamics plus N-body Adaptive Refinement Tree (ART) code (Kravtsov 1999, 2003). The zoom-in technique (Klypin et al. 2001) applied here allows us to resolve the formation of single galaxies consistently in their full cosmological context.

We analyzed the VELA simulation suite (Ceverino et al. 2014; Zolotov et al. 2015), which was created to compliment the HST CANDELS survey (Barro et al. 2013, 2014). The hydrodynamic code used to simulate these galaxies incorporates prescriptions for star formation, stellar feedback, Type II and Ia supernova metal enrichment, radiation pressure, self-consistent advection of metals, and metallicity-dependent cooling and photoionization heating due to a cosmological ultraviolet background. Our simulations have a feedback model, named RadPre_LS_IR (Ceverino et al. 2014), that differs from previous studies (Zolotov et al. 2015). This model includes radiation pressure from infrared photons, as well as photoheating/photoionization around young and massive stars. Further details regarding the various models included in these simulations can be found in Ceverino & Klypin (2009) and Ceverino et al. (2014).

These simulations resulted in a maximum spatial resolution of 17 pc, a dark matter particle mass of 8 × 10⁴ M_⊙, and a minimum stellar particle mass of 10³ M_⊙. The high resolution implemented in the VELA simulations allows us to resolve the regime in which stellar feedback overcomes the radiative cooling (Ceverino & Klypin 2009), which results in naturally produced galactic scale outflows (Ceverino et al. 2010, 2016). Thus, galaxy formation proceeds in a more realistic way through a combination of cold flow accretion, mergers, and galaxy outflows.

Here we select a subsample of the VELA galaxies that (a) were evolved to the lowest redshift of z = 1, (b) did not experience a major merger near z = 1, and (c) have a virial mass range of log M_vir = 11.3–12 (see Table 4 for halo virial quantities). The selection resulted in eight galaxies having an average log M_vir = 11.7 ± 0.2 M_⊙ and log M_* = 10.5 ± 0.3 M_⊙.

Table 4.  Properties of z = 1 VELA Galaxies

VELA	log(${M}_{\mathrm{vir}}/$ ${M}_{\odot }$)	log(${M}_{* }/$ ${M}_{\odot }$)	Rvir
Galaxy			(kpc)
21	12.0	10.9	151
22	11.8	10.7	133
23	11.7	10.4	118
25	11.5	10.2	103
26	11.6	10.4	112
27	11.6	10.3	110
28	11.3	9.9	92
29	12.0	10.6	146

Download table as:  ASCIITypeset image

We employed the HARTRATE photo+collisional ionization code (Churchill et al. 2014) that is optimally designed to model optically thin gas with no ionization structure. For the vast majority of the CGM, including the O vi column densities and impact parameters studied here, this is a safe assumption. The consequence of not including any optical depth considerations is that HARTRATE may underpredict ions that typically reside in optically thick conditions (such as Mg ii); however, this is not a concern here since this assumption only breaks down close to central galaxies or near satellite galaxies. A proper treatment of the radiation field would require computationally intensive full radiative transfer computations, which is beyond the scope of this work.

In summary, HARTRATE incorporates photoionization, direct collisional ionization, Auger ionization, excitation-autoionization, photo-recombination, high/low-temperature dielectronic recombination, charge transfer ionization by H⁺, and charge transfer recombination by H⁰ and He⁰. HARTRATE uses solar abundance mass fractions (Asplund et al. 2009; Draine 2011); a Haardt & Madau (2012) ionizing spectrum is used for the ultraviolet background and assumes ionization equilibrium. The cosmological simulations provide HARTRATE with the hydrogen number density, kinetic temperature, and gas metallicity (i.e., Type II and Ia supernova yields). The outputs from HARTRATE include the electron density, the ionization and recombination rate coefficients, ionization fractions, and the number densities for all ionic species up to zinc. The software has been applied successfully in previous works (Churchill et al. 2012, 2015; Kacprzak et al. 2012); see Churchill et al. (2014) for details on the code and its successful comparisons to CLOUDY.

The methodology of producing mock observations of quasar sightlines through the cosmological simulations is described in detail in Churchill et al. (2015) and Vander Vliet (2017). The outputs from HARTRATE are applied to Mockspec, which performed the mock quasar absorption analysis. Mockspec is publicly available in a GitHub repository.⁷ We ran HARTRATE on a smaller box size of 6R_vir along a side centered on the dark matter halo of the host galaxy and drew 1000 lines of sight within a maximum impact parameter of 1.5R_vir.

Absorption spectra with the instrumental and noise characteristics are generated assuming that each cell gives rise to a Voigt profile at its line-of-sight redshift. The mock quasar sightline is then objectively analyzed for absorption above the EW threshold of 0.02 Å, which corresponds to $\mathrm{log}N({\rm{O}}\,{\rm{VI}})=13.55$ cm⁻² for b = 10 km s⁻¹. The optical-depth-weighted median redshifts, rest-frame EWs and velocity widths, and column densities are then measured from the spectra (see Churchill & Vogt 2001). The velocity zero-point of the simulated absorption lines is set to the line-of-sight velocity of the simulated galaxy (center of mass of the stars). For this analysis, all eight simulated galaxies are analyzed with the disk appearing edge-on to the observer. The galaxy inclination is determined relative to the angular moment vector of cold gas (T < 10⁴ K) within 1/10 of R_vir. The systemic velocity of the galaxy is determined by the dark matter particles within the halo virial radius.

To examine the spatial and kinematic properties of gas giving rise to O vi absorption, we identify O vi absorbing gas cells along each sightline as those that contribute to detected absorption in the simulated spectra. The gas cells along the sightline are sorted into decreasing column density, and the lowest are systematically removed until the noiseless spectrum created by the remaining cells has an EW that is 95% of the EW of the original spectrum.

In Figure 6, we show the O vi column density distribution from the simulations (gray) and from the observations of Kacprzak et al. (2015) (red). We note an anticorrelation for both the observations and simulations between the column density and impact parameter. There is overlap between the simulated and observed column densities, while also being consistent with previous works using simulations (Hummels et al. 2013; Ford et al. 2016; Liang et al. 2016; Oppenheimer et al. 2016; Gutcke et al. 2017; Suresh et al. 2017). Although the simulations show a larger degree of scatter, which could be driven by galaxy inclination, etc., they can still provide useful insight into the kinematics driving the existence of O vi systems. We will explore this scatter and offsets between observations and simulations in an upcoming paper.

Zoom In Zoom Out Reset image size

Figure 7 shows the median O vi column density distribution for sightlines through the simulations for a single example galaxy of VELA 27. Only the cells contributing the O vi absorption (as described in the previous section) are shown. The coordinate system for the example galaxy is defined so that the disk lies in the xy-plane with the angular moment vector of cold gas along the positive z-axis. The black circle indicates the virial radius. A somewhat spherical O vi halo is present within ∼40–50 kpc of the galaxy center and has almost unity covering fraction. This spherical halo around the host galaxy has column densities ranging between log N(O vi) = 12.5 and 14.

Zoom In Zoom Out Reset image size

Beyond 50 kpc are possibly three thick filaments responsible for producing the high impact parameter absorption with column densities decreasing to log N(O vi) = 12.5–11. These two features, halos and streams, are seen in all of our simulated galaxies. Note that in this particular example galaxy the filaments are not coplaner and tend to be in different locations for all galaxies. We will explore the spatial distribution of O vi in an upcoming paper. Next, we examine whether these structures in the simulations are able to reproduce the typical absorption profiles and kinematics seen in our observations.

Figure 8 shows the same O vi gas cells contributing to the absorption profiles as seen in Figure 7, but now color-coded as a function of velocity in spherical coordinates v_r, v_θ, and v_ϕ. Here the median velocity of all the O vi cells contributing to the absorption along each projection of the sightlines is shown.

Zoom In Zoom Out Reset image size

The top panel has the radial velocity component showing what speeds the O vi gas is traveling directly away from or toward the center of the galaxy. It can be clearly seen that there is significant radial inflow toward the galaxy center along the filament structures. In this particular example, the inflowing gas appears to have a roughly constant velocity ranging between −150 and −200 km s⁻¹, with potentially an increase toward the galaxy center. The central part of the galaxy halo has a component that exhibits slower inflow velocities of 0–100 km s⁻¹ and sits both near and outside of the filaments. Most of the gas near the galaxy averages along the line of sight is close to the systemic velocity. We see only a few gas cells in this example that have positive, radially outflowing velocities ranging from 0 to 100 km s⁻¹.

The middle panel shows v_θ, which is the rotation velocity, where gas corotating with the galaxy has positive speeds and gas counterrotating with the galaxy has negative velocities. In the inner 50 kpc, the gas is rotating in the same direction as the galaxy having velocities between 50 and 100 km s⁻¹. This corotating gas appears to be in the same plane as the edge-on disk galaxy, suggesting some connection between the O vi and disk gas. There is also some gas within 50 kpc that is near the systemic velocity and some counterrotating with speeds <50 km s⁻¹. Beyond 50 kpc, most of the gas shows little sign of rotation, while some gas is counterrotating with a range of velocities from 0 to −150 km s⁻¹. The dominant velocity component outside of 50 kpc is the radial component.

The bottom panel shows v_ϕ, which is the rate of change of the angle between the vector to the gas cell and the z-axis, which is aligned with the galaxy's angular momentum vector. In the central region, we see positive velocities that decrease closer to systemic velocity with increasing impact parameter. There are some negative velocities out in the filaments as well.

We next explore the eight simulations in a statistical sense in order to determine general kinematic trends and origins of the O vi CGM.

The top panel of Figure 9 shows the mean stacked Voigt profile fits to the O vi that is located along the galaxy major (Φ < 25°) and minor (Φ < 33°) axes for our observations. Note that both have similar kinematic shape and are centered near the galaxy systemic velocity. The major-axis gas is offset by 2.5 km s⁻¹ from the galaxy systemic velocity, while the minor-axis gas is offset by 28.0 km s⁻¹ from the galaxy systemic velocity. This implies that there are no strong kinematic signatures present if outflows and accretion are traced by O vi gas, or outflow and accretion signatures could be hidden by a larger diffuse collection of O vi within the halo at similar velocities.

Zoom In Zoom Out Reset image size

To compare our observations to the simulations, we select all absorption systems from the eight simulated galaxies that have an EW larger than 0.2 Å, which is roughly the observational limit of our sample. For the simulations, major-axis gas is defined as having an azimuthal angle less than 30°, while minor-axis gas has an azimuthal angle greater than 40°. These absorption systems were then combined to provide the mean stacked spectra shown in Figure 9. We note that the optical depths and the velocity spread between the simulations and observations are similar, with some differences with the kinematic shape of the profile.

The O vi found near the major axis in the simulations exhibits a possible bimodal distribution with the bulk of the absorption residing near 100–125 km s⁻¹ on each side of the galaxy systemic velocity. This signature is reminiscent of corotation/accretion predictions (Stewart et al. 2011, 2013, 2017; Danovich et al. 2015). O vi found near the minor axis in the simulations exhibits an offset of ∼50 km s⁻¹ from the galaxy systemic velocity but has a similar velocity spread to the observations. The simulated mean stacked spectra do show some hints of kinematic structures, such as rotation along the major axis, which does differ from our observations. This could be due to only having 10 sightlines from our observations, or differences due to inclination angle effects. We next examine the typical O vi velocities to determine what is driving the O vi kinematic distribution within the simulations.

Figure 10 shows the median O vi cell velocities averaged over the eight simulations shown for gas along the major (red) and minor (blue) axis. The left panel shows the radial velocity component for major- and minor-axis gas along with the standard error in the mean. Gas along the major axis of the galaxy appears to inflow toward the galaxy at high velocities at high impact parameter and slows to the galaxy systemic velocity as it approaches the galaxy center. The largest deceleration occurs within 50 kpc, reducing in speed from −50 to 0 km s⁻¹. Thus, both Figures 8 and 10 indicate that O vi gas does inflow along filaments and decelerates as it approaches the galaxy.

Zoom In Zoom Out Reset image size

On the other hand, minor-axis O vi is outflowing out to about 50 kpc, and then it decelerates and falls back toward the galaxy. The minor-axis gas has similar radial velocities to the major-axis gas beyond 75 kpc, which would make it difficult to identify the difference between accreting and reaccreted gas. Thus, outflows traced by O vi only influence the CGM out to 50 kpc for a Milky-Way-like galaxy, and recycled outflows, which are a common prediction from simulations as an origin of O vi gas, dominate at higher impact parameters. This is consistent with the toy outflow models in Section 3.3, indicating that if the gas is originating from outflows, the gas has to be decelerating and possibly falling back to the galaxy. Furthermore, our minor-axis observational sample contains three galaxies with impact parameters less than 50 kpc. In those three cases (J1241, J1555, J2253; z_gal = 0.1537) the O vi resides to one side of the galaxy systemic velocity, so it is possible that they exhibit signatures of gas outflows.

The middle panel shows the rotational angular velocity, v_θ, where positive velocities indicate that gas is rotating in the same direction as the galaxy. The major-axis gas is rotating in the same direction as the galaxy as it infalls toward the disk. The rotation velocity component increases within 100 kpc and becomes the dominant velocity component near the galaxy. Thus, we should see clear signatures of corotation in our observations. The minor-axis gas may be rotating in a similar direction within 25 kpc, but then it scatters around zero, showing little sign of following the direction of galaxy rotation.

The right panel shows the polar velocity, which is the rate of change of the angle between the vector to the cell and the z-axis, which is aligned with the galaxy's angular momentum vector. This is the lowest velocity component for the major-axis gas, showing that this gas is primarily infalling, corotating, and not mixing very much azimuthally. The minor-axis gas has roughly zero polar velocity within 50 kpc and beyond 125 kpc. Between 50 and 125 kpc, the gas begins to have negative velocities. This occurs over the same impact parameter range where the radial velocity of the minor-axis gas transitions from outflowing accelerating velocities to decelerating and accreting velocities, indicating a change in the behavior of the kinematics of minor-axis O vi gas. This is a signature of the O vi returning back to the disk plane of the galaxy. Overall the dominant minor-axis velocity component is radial, be it outflowing or accreting.