Absorption Line Search Through Three Local Group Dwarf Galaxy Halos

Dwarf galaxies are missing nearly all of their baryons and metals from the stellar disk, presumed to be in a bound halo or expelled beyond the virial radius. The virial temperature for galaxies with $M_{\rm h} \sim 10^9 - 10^{10}$ $M_{\odot}$ is similar to the collisional ionization equilibrium temperature for the C IV ion. We searched for UV absorption from C IV in six sightlines toward three dwarf galaxies in the anti-M31 direction and at the periphery of the Local Group ($D \approx$ 1.3 Mpc; Sextans A, Sextans B, and NGC 3109). The C IV doublet is detected in only one of six sightlines, toward Sextans A, with $\log N({\rm C IV})$ = $13.06 \pm 0.08$. This is consistent with our gaseous halo models, where the halo gas mass is determined by the cooling rate, feedback, and the star formation rate; the inclusion of photoionization is an essential ingredient. This model can also reproduce the higher detection rate of O VI absorption in other dwarf samples (beyond the Local Group), and with C IV only detectable within $\sim 0.5R_{\rm vir}$.


INTRODUCTION
Several lines of studies reveal that galaxies have lost most of their baryons and most of their metals. The "missing baryons" problem is the finding that the mass of stars and cold disk gas is only about one quarter of the cosmic baryon mass expected for the dynamical mass of a galaxy (McGaugh et al. 2010). The fraction of baryons declines toward lower mass galaxies, where dwarfs may only account for only 5% of their baryons. Closely related is the "missing metals" problem, where about 70 − 80% of the metals formed by galaxies are present in the stars and cold gas (Peeples et al. 2014).
A likely solution is that the baryons and metals exist as a gaseous halo surrounding the galaxy, possibly extending beyond the virial radius. This requires strong feedback due to supernovae and active galactic nucleus (AGN), which is predicted to have a profound effect on the gaseous halo (e.g., Oppenheimer 2018; Davies et al. 2020. Many searches focus on the more massive galaxies (M h > 10 10 M ), which should have more extended halos, and there have been a number of successes (e.g., Stocke et al. 2013;Werk et al. 2014). In massive galax-Corresponding author: Joel N. Bregman jbregman@umich.edu ies, the ultraviolet (UV) absorption lines occur at temperatures below virial, so the gas that they trace is not the volume filling component.
The high ionization UV lines can trace the volumefilling medium in lower mass galaxies, the targets of this program. The C IV ion can produce strong UV absorption lines and typically is found in gas at 6 × 10 4 K in collisional ionization equilibrium, which corresponds to the virial temperature for a M halo ∼ 10 9 M system. Such systems correspond to dwarf galaxies that are less massive than the Small Magellan Cloud (SMC) but more massive than the very faint dwarf galaxies in the Local Group. It might be possible to detect such hot halos or winds around dwarf galaxies using ions like C IV or Si IV. In this study, we search for evidence of these absorption lines around three dwarf galaxies in the Local Group, NGC 3109, Sextans A, and Sextans B (Fig. 1). These galaxies were chosen because they are far from the Milky Way and M31 (D = 1.33 Mpc), and they are in the anti-M31 direction. In this direction, the velocity difference between the Sun and these galaxies is 300 − 400 km s −1 , largely due to the infall of the Milky Way toward M31. This provides enough velocity separation that one can hope to identify absorption as being due to the dwarf halos.

The NGC 3109 Association of Dwarfs
NGC 3109 is the most massive of the galaxies in this "group", with M * + M (H I) = 6.7 × 10 8 M , or about 10 −2 of the equivalent Milky Way value, although with the important difference that the H I mass is significantly larger than the stellar mass (Carignan et al. 2013), as M * ≈ 8 × 10 7 M . It is classified as a Magellanic-type spiral that is close to being edge-on, and with a rotation curve that is flattening by 8.25 kpc (their last data point), with an implied dynamical mass of at least 8.6 × 10 9 M . The H I gas is more extended than the stellar disk, which is typical of a field spiral, and it has ongoing star formation at a rate of ∼ 5 × 10 −3 M yr −1 .
The dwarf irregular galaxies Sextans A and Sextans B are similar in that they have the same luminosity and stellar mass (M * ≈ 4 × 10 7 M ) and dynamical mass (1 × 10 9 M ; Bellazzini et al. 2014). In Sextans A, the H I gas is less extended than the asymmetric stellar light distribution and has M (H I) = 6.2 × 10 7 M , which is slightly larger than the stellar mass. In Sextans B, the H I and stellar light are nearly identical in size and shape, where M (H I) = 4.1 × 10 7 M , about equal to the stellar mass. Star formation is active in both systems, with similar rates during the past Gyr of ∼ 2 × 10 −3 M yr −1 (Weisz et al. 2011).
NGC 3109, Sextans A, and Sextans B are at nearly the same distance and part of the sky, which might be coincidental, but Tully et al. (2006) showed that many, if not most of the dwarfs in the Local Group lie in associations, and sometimes in well-defined planes (Libeskind et al. 2015). The association containing NGC 3109, Sextans A, and Sextans B has at least two other confirmed and five other possible nearby members (Tully et al. 2006;Bellazzini et al. 2013). These galaxies are separated by distances (250 − 500 kpc) that are large compared to their virial radii (40 − 80 kpc; Fig. 1), so we can view them as each being primary galaxies.
This group of galaxies may have passed closer to the center of the Local Group in the past, but without a strong interaction with an ambient medium, as ram pressure stripping has not made these galaxies H I-poor (Putman et al. 2021). The orbital period is comparable to the age of the Universe, although the dynamical models are not in perfect agreement with the velocity and position of these dwarfs (Banik & Zhao 2017;Peebles 2017). This may be due to the gravitational influence of the Virgo Cluster (Banik & Zhao 2017), as these galaxies are nearly in the Hubble flow. Regardless of these details, these galaxies have spent more than a Gyr in a relatively poor environment, not near any massive galaxy. A galactic wind or hot halo could have established itself in this period of relative isolation.

DATA REDUCTION AND RESULTS
In this study, there are six QSO sight lines around three dwarf galaxies (NGC 3109, Sextans A, and Sextans B) within the local group. These observations are obtained by two HST programs (13347 and 11524), which is summarized in Table 2.
The data reduction follows Wakker et al. (2015) and Qu et al. (2019), and we briefly summarize the reduction steps here. First, we co-add the gross counts obtained from individual exposures, which are corrected for the background and the fixed pattern noise. Then, the net count rates are calculated from the gross counts by dividing the co-added effective exposure times. The co-added noise is the Poisson noise derived from the total counts, which is found to be better matched with the measured noise of the co-added spectrum than the co-added noise (Wakker et al. 2015). The final coadded spectrum is in the heliospheric frame. The S/N ratios of the co-added spectra are also summarized in Table 2.
There is a known instrumental shift issue for the wavelength solution calculated from CALCOS for the COS spectrum (Wakker et al. 2015). Here, we performed a wavelength calibration for the co-added spectra, which employs the H I 21 cm lines as references for UV absorption lines. The H I 21 cm emission lines are extracted from the Leiden/Argentine/Bonn (LAB) Survey of Galactic H I with an effective beam size of 0.2 • around the QSO sight lines (Kalberla et al. 2005). The Galactic absorption lines of low ionization state ions (e.g., Si II  (2) and (3) Galactic coordinates of the galaxy; (4) distance to the galaxy; (5) projected heliospheric velocity; (6) Absolute K band magnitude; (7) background AGN; (8) and (9) Galactic coordinates of the AGN; (10) redshift of the AGN; (11) FUV magnitude of the AGN; (12) projected distance between the AGN and the galaxy.  (3) and (4)  and C II) are assumed to have the same velocity line centroid as the H I 21 cm lines, and a set of velocity shifts are obtained over the entire spectrum. Then, we fit polynomial functions to the obtained velocity shifts simultaneously for the four spectrum segments in both G130M and G160M. In most cases, we only use a constant or linear functions to correct the wavelength, and only the segment B of G130M of PG 1001+054 use a second order polynomial. The co-added spectrum is binned by three pixels for fitting and line identification, yielding the bin width of ∆λ = 0.02991Å and 0.03669Å for G130M and G160M, respectively. Most lines are non-detection. The equivalent widths (EWs) and uncertainties are calculated over a typical width of 60 km s −1 , which is about three times the typical b values (20 km s −1 ) measured for cool-warm CGM. The only detection is C IV of PG 1011-040 for the Sextans A at heliocentric v = 294 km s −1 (Fig. 2), and −30 km s −1 from the systemic velocity of Sextans A. We fit the Voigt profile for this C IV detection, obtaining the column density of log N = 13.04 ± 0.08 and b = 9.9 ± 5.3 km s −1 . The 2 σ upper limits of the column densities, are calculated from EW uncertainties by assuming b = 20 km −1 for other none-detection (Table  3).
In this sight line of PG 1011-040, there is also a possible Si III λ1206.5Å feature at v helio = 280 km s −1 , which is different in velocity from the 294 km s −1 C IV system by 3σ. We do not consider it as an absorption feature associated with Sextans A for two reasons. First, Si III only has one transition at 1206.5Å in the HST/COS coverage, and no other ions show absorption features at a similar velocity. Thus, it may be contamination from intervening absorption systems beyond the Local Group. Second, this feature is 44 km s −1 away from the bulk velocity from Sextans A, which is beyond the rotation velocity of the galaxy (≈ 30 km s −1 ; Namumba et al. 2018).  (3) impact parameter in kpc; (4) -(9) log of the column densities of ions, where the column is in units of cm −2 .

DISCUSSION AND CONCLUSIONS
Our only detection is in the sight line closest to Sextans A, where the separation is 55 , or 21 kpc, well within the virial radius, ≈ 44 kpc (Fig. 1). This can be compared to the outermost radius at which the H I emission is detected in LITTLE THINGS (the Very Large Array, VLA) and by the Square Kilometre Array (SKA) pathfinder KAT-7 (Hunter et al. 2012;Namumba et al. 2018). The two observations yield the same radial profile and have a limiting 3σ H I column density for the VLA observation is 2 × 10 18 cm −2 and a value of 5.8 × 10 18 cm −2 for the KAT-7 observation. The H I is detected to a radius of 3.5 kpc (9 ), which is six time smaller than the sight line-galaxy separation. This indicates that the absorption is more likely due to the gaseous halo around Sextans A rather than an extension of the disk.

Comparison to Other Dwarf Surveys
There are four studies of dwarf galaxies beyond the Local Group that can be used for comparison. They differ somewhat in selection criteria and redshift space, and for M * similar to our galaxies, are either at z < 0.02  (Stocke et al. 2013;Bordoloi et al. 2014;Burchett et al. 2016) or z ≈ 0.2 (Johnson et al. 2017). The C IV absorption detection rate is similar between these surveys, and for 7.5 < log M * < 9, the detection rate is 3/37 galaxies, summed over the four surveys. For the detected systems, the impact parameter is < 0.5R vir , and a similar result is found for more massive dwarfs, 9 < log M * < 10 Burchett et al. (2016). We acknowledge differences between the sample but find that the results of these surveys are consistent with the absorption results in this work, where 1/6 sightlines have a C IV detection, and that sight line is the closest to the galaxy (Sextans A). The probability of having one detection for six sightlines is about 50%, assuming a Poisson probability of 3/37 per sightline, so this our detection rate is in agreement with the surveys.
It is worth noting that the absorption line detection rate is larger for H I and O VI (ions that we cannot observe in our sample). Johnson et al. (2017) finds that H I is detected in nearly every galaxy (17/18) and that O VI is detected in one-third of the galaxies (6/18). We return to this result for O VI when discussing absorption models for dwarf galaxies (below).
One can also consider some of the observations toward other dwarfs in the Local Group. One of the problems with such studies, including ours, is that Galactic gas comes in various forms (e.g., high-velocity cloud, intermediate-velocity cloud, and Magellanic Stream), which covers a range of velocities (Richter et al. 2017). This can lead to a confusion problem that one tries to resolve by using velocity separation from the known gaseous features in the halo.
The dwarfs IC 1613 and WLM are in the same part of the sky as the Magellanic Stream and within 30 − 40 • of M33 and M31. In WLM, Zheng et al. (2019) find absorption components that include C IV at velocities of −150 and −220 km s −1 . These velocities can be compared to the systemic velocity of WLM (−132 km s −1 ), and the Magellanic Stream, where the mean velocity is < −190 km s −1 . They tentatively assign the −150 km s −1 system to the halo of WLM.
The dwarf IC 1613 has four sightlines toward hot stars in the dwarf and six more from AGN sight lines (Zheng et al. 2020). This is more massive than WML, having M * ≈ 1 × 10 8 M , M (H I) = 0.65 × 10 8 M , and M h ≈ 4 × 10 10 M , and at a velocity shift of −232 km s −1 . Zheng et al. (2020) identify absorption, typically within 100 km s −1 of the systemic velocity of IC 1613 for all ten sightlines. This 100% success rate makes IC 1613 quite unusual relative to other dwarfs. They discuss whether some of these features might be associated with the Magellanic Stream, but conclude that IC 1613 is the most likely absorption line host.
We also consider the contamination from unassociated Galactic gas for Sextans A. From the survey of 270 sightlines, Richter et al. (2017) shows that sightlines in this general direction (200 • < l < 300 • , 30 • < b < 60 • have absorption at velocities of 100 − 200 km s −1 , which we find in our sightlines as well. However, it is impossible to be certain that a small amount of halo gas, unrelated to Sextans A, is responsible for the absorption at 294 km s −1 .

Comparison to Gaseous Halo Models
One can try to estimate whether a gaseous halo around a dwarf galaxy would have enough column density in metals to be detectable. The upper limit in the column is N 200 = n 200 R 200 , where n 200 is 200 times the critical baryon density of the universe at z = 0. For our dwarfs, this yields N 200 = 5 − 10 × 10 18 cm −2 , and for gas at 0.1 Solar metallicity, and an ionization fraction of 0.2 (near the peak), the C IV column density would be N (C IV) 200 = 3 − 5 × 10 13 cm −2 , which is somewhat larger than our upper limits for C IV of 1.3 − 2 × 10 13 cm −2 , or the single C IV detection of 1.48 ± 0.30 × 10 13 cm −2 . However, agreement would occur with a somewhat lower C IV ionization fraction, which motivates the need for more realistic models.
A more accurate and realistic estimate is possible by considering the processes likely to occur in the halos of galaxies. This includes radiative cooling, feedback, and photoionization from the ambient metagalactic radiation field. These processes were considered for a range of galaxy masses in Qu & Bregman (2018a) and Qu & Bregman (2018b), which consider several models, including those with collisional ionization equilibrium and   Figure 4. Left panel: the equivalent width of C IV λ1548 (inÅ) versus the fractional virial radius for galaxies with two stellar masses and at two redshifts. The two distributions by stellar mass (at fixed redshift) are similar, with a sharp decline beyond ≈ 0.8Rvir at z = 0 and ≈ 1.2Rvir at z = 0.2. The difference is mainly due to the increased star formation rate at higher redshift. Right panel: the equivalent width of C IV λ1548 (inÅ) vs the projected radius in kpc for the three z = 0 galaxies in this study, which have logM * /M = 7.6 − 8.6. The only detection is for Sextans A (logM * /M = 7.6 at the smallest radius; downward triangles are upper limits. The points for Johnson et al. (2017) are in the range logM * /M = 7.7 − 9.2 and at z ≈ 0.2. The model seems to be consistent with the data for logM * /M ≈ 8 but might over-predict the columns for logM * /M ≈ 9.
photionization equilibrium. The most realistic model is where phototionization (Haardt & Madau 2012) is included in the ionic distribution and where the net cooling rate of the gaseous halo equals the star formation rate in the galaxy. For star formation, we use a mean value for a galaxy of mass M h and a typical feedback rate (details are given in the above papers). A multitemperature halo gas results from the cooling and the feedback, which we denote as the TPIE model (defined in Qu & Bregman 2018a). Here we extend the mass of the halos to values lower than previously published, along with column densities for both O VI and C IV. In the TCIE model (collisional ionization equilibrium in the multitemperature gas), the O VI column density turns sharply downward for logM h < 11.4, as the virial temperature falls below the peak O VI ionization fraction. However, in the TPIE model, that turndown at lower M h does not occur because photoionization raises the fractional ionization (Fig. 3). The result is that the C IV column density varies by less than a factor of two for 7.5 < logM * /M < 10.5 (z < 0.2), and rarely rises above a column of 10 13 cm −2 . This is similar for the O VI ion, although the column density is larger by a factor of three to six. There is scatter in these column densities, as galaxies differ in their star formation rate and halo mass. In general, the model suggests that the typical N (C IV) column density lies below the HST/COS detection limit in these surveys, while N (O VI) is close to the detection threshold.
We also calculated the equivalent widths of the C IV λ1548 as a function of with radius (Fig. 4). There is a decline in the equivalent width with radius, typically by a factor of three from 0.1 − 0.5R vir , and another factor of two to five from 0.5 − 1.0R vir , depending on mass and redshift (Fig. 4). This helps to explain why absorption detection are rarely found beyond ≈ 0.5 R vir , although more detectable absorption lines were expected for the dwarfs with logM * /M ≈ 9.
To summarize, the TPIE models produce predictions for the column density of C IV and O VI that reproduce some of the features found in studies of absorption of dwarfs beyond the Local Group. This suggests that dwarf galaxies contain extended gaseous halos, provided they are not in environments where their halo has been stripped by the ambient hot medium of a more massive galaxy.