The Metallicities of Five Small High-velocity Clouds

High-velocity clouds (HVCs) are multiphase gas structures whose velocities (∣v LSR∣ ≥ 100 km s−1) are too high to be explained by Galactic disk rotation. While large HVCs are well characterized, compact and small HVCs (with H i angular sizes of a few degrees) are poorly understood. Possible origins for such small clouds include Milky Way (MW) halo gas or fragments of the Magellanic System, but neither their origin nor their connection to the MW halo has been confirmed. We use new Hubble Space Telescope/Cosmic Origins Spectrograph UV spectra and Green Bank Telescope H i spectra to measure the metallicities of five small HVCs in the southern Galactic sky projected near the Magellanic System. We build a set of distance-dependent Cloudy photoionization models for each cloud and calculate their ionization-corrected metallicities. All five small HVCs have oxygen metallicities ≤0.17 Z ⊙, indicating they do not originate in the disk of the MW. Two of the five have metallicities of 0.16–0.17 Z ⊙, similar to the Magellanic Stream, suggesting these clouds are fragments of the Magellanic System. The remaining three clouds have much lower metallicities of 0.02–0.04 Z ⊙. While the origin of these low-metallicity clouds is unclear, they could be gaseous minihalos or gas stripped from dwarf galaxies by ram pressure or tidal interactions. These results suggest that small HVCs do not all reside in the inner MW halo or the Magellanic System, but instead can trace more distant structures.


INTRODUCTION
High-velocity clouds (HVCs) are multi-phase gas clouds observed at velocities too large to be accounted for by disk rotation, typically |v LSR | ≥ 100 km s −1 (Wakker & van Woerden 1997).HVCs play an integral role in regulating the gaseous ecosystem of the Milky Way (MW) and its star formation rate (Putman et al. 2012;Richter et al. 2017;Fox et al. 2019).Their inflow to the MW disk provides new fuel for future star formation, while their outflow reduces the available gas supply.Fox et al. (2019) and French et al. (2021) use HVC UV data to show that the MW is currently in an inflow-dominated phase, with an HVC inflow rate of ≳0.53±0.23 M ⊙ yr −1 .While this inflow rate is not sufficient to sustain the current star formation rate of 1.7 ± 0.7 M ⊙ yr −1 (Chomiuk & Povich 2011;Licquia & Newman 2015), HVCs still play an important role in replenishing the MW gaseous disk with fresh gas for future generations of star formation.
A well-known source of HVCs in the southern Galactic hemisphere is the Magellanic System.The System contains the Large and Small Magellanic Clouds (LMC and SMC), the Magellanic Stream of stripped gas trailing the Clouds, the Magellanic Bridge of gas connecting the Clouds, and a gaseous Leading Arm in front of the Clouds (see Putman et al. 1998Putman et al. , 2003;;Brüns et al. 2005;Nidever et al. 2008Nidever et al. , 2010;;D'Onghia & Fox 2016, and references therein).HVCs belonging to the Magellanic System have been identified by their projected positions, position-velocity diagrams, and low metallicities of typically 0.10 − 0.17 Z ⊙ (Lu et al. 1998;Gibson et al. 2001;Fox et al. 2010Fox et al. , 2013Fox et al. , 2018;;Richter et al. 2013Richter et al. , 2018)); albeit the metallicity of the LMC filament of the Stream is ≈ 0.50 Z ⊙ (Gibson et al. 2000;Richter et al. 2013).While the Magellanic System dominates the HVC population in the southern Galactic hemisphere, there are a significant number of HVCs that are not clearly associated with the Magellanic System and whose origin is unknown (Richter et al. 2001;Putman et al. 2012;Westmeier 2017).
A poorly-understood subset of HVCs are the Compact HVCs (CHVCs), defined as isolated clouds with H I diameters on the sky of < 2 • (Braun & Burton 1999;de Heij et al. 2002a,b,c;Putman et al. 2002;Saul et al. 2012).Many small HVCs also exist that are not isolated or are slightly larger than 2 • ; these are not formally considered CHVCs, but their origin is equally unknown.CHVCs have been extensively studied in H I, while infrared searches for stellar components in CHVCs have not been successful, making it difficult to put strong constraints their distances (Braun & Burton 1999;Hopp et al. 2003Hopp et al. , 2006)).Distance estimates of CHVCs from H I observations based on kinematics, thermal pressure, comparisons to known dwarf galaxies, the spatial distribution of CHVCs, and the virial theorem yield distances of ∼100-850 kpc, placing the CHVCs firmly within the Local Group (Braun & Burton 2000;Brüns et al. 2001;Burton et al. 2001;Sternberg et al. 2002;de Heij et al. 2002c;Maloney & Putman 2003;Westmeier et al. 2005;Pisano et al. 2004Pisano et al. , 2007)).
Compact and small HVCs have many possible origins, including: (1) gas ejected from the Galactic disk in an energetic outflow, (2) extragalactic gas infalling through the Galactic halo, (3) fragments of the Magellanic System, (4) ultra-faint Local Group dwarf galaxies whose stars are yet to be detected or mini-halos, small darkmatter halos with gas and no star formation, (5) gas stripped from Local Group dwarf galaxies via tidal interactions or ram pressure, and (6) intergalactic gas (Oort 1966;Giovanelli 1978;Bregman 1980;Braun & Burton 1999;Blitz et al. 1999;Brüns et al. 2000;Brüns et al. 2001;Sternberg et al. 2002;de Heij et al. 2002c;Pisano et al. 2004Pisano et al. , 2007;;Putman 2006;Putman et al. 2012).For example, five CHVCs have been identified by Nidever et al. (2010) as Magellanic Stream (MS) components based on their continuity in position-velocity diagrams with the MS.However, CHVCs are found across the sky and therefore cannot always be attributed to Magellanic gas (Putman et al. 2012;Moss et al. 2017).
UV absorption-line studies can be used to measure the ionized gas content, metallicities, and dust depletion levels of compact and small HVCs and therefore constrain their origin.Until now, UV absorption studies have been conducted for only two CHVCs: CHVC 224.0−83.4−197,by Sembach et al. (2002), Richter et al. (2009), and Kumari et al. (2015) and CHVC 125+41−207, by Bowen & Blades (1993) and Braun & Burton (2000).  is located in the southern Galactic hemisphere ∼10 • from the H I MS, and is probed by the sightline to the QSO Ton S210.Kumari et al. (2015) determined a low metallicity of [O/H] = −1.12±0.22(0.076 Z ⊙ ) for this CHVC and concluded it is likely a fragment of the MS.CHVC 125+41−207 is located in the northern Galactic hemisphere and has a measured metallicity of 0.04-0.07Z ⊙ .Braun & Burton (2000) suggest that this measurement along with H I kinematics of CHVC 125+41−207 imply a Local Group dark-matter-dominated self-gravitating object, possibly a low surface brightness dwarf galaxy.These are valuable measurements, but more a larger sample of metallicities is needed to draw general conclusions.
In this paper, we present the first systematic study of the metallicities of small HVCs using a sample of five southern Galactic hemisphere clouds identified from the Galactic All-Sky Survey (GASS) (McClure-Griffiths et al. 2009;Moss et al. 2013Moss et al. , 2017)).The UV and H I observations and data reduction are presented in Section 2. In Section 3 we present the calculations for each HVC's ionization-corrected gas-phase abundances and dust depletion levels.We discuss potential explanations for the HVC metal abundances in Section 4. We then summarize our results in Section 5.
c Local Standard of Rest (LSR) velocity of the HVC, measured by Moss et al. (2013).
d HVC angular size in Right Ascension and Declination within the H I contours shown in Figure 2. Burton (1999), because two have sizes just above the 2 • threshold, and two others are not isolated, which is a secondary factor in identifying CHVCs; instead, we refer to the five clouds as "small HVCs" throughout this analysis.The HVCs all have identifications in the original HIPASS catalog of HVCs (Putman et al. 2002) as well as GASS identifications.Each of the five background AGN lies directly behind their associated HVC H I emission and less than 1 • from the peak GASS H I flux (see Figure 2).See Table 1 for basic information on each sight line and HVC, including their GASS and HIPASS identifications.For brevity, we adopt abbreviated names for the following background AGN: HE 0027 = HE 0027-3118, IRAS 0459 = IRAS 04596-2257, and UVQS J0110 = UVQS J011054.99-154540.2.

UV Observations
Each target AGN was observed with the Hubble Space Telescope Cosmic Origins Spectrograph (HST/COS) under Cycle 27 Program 15887 (PI: A. Fox).The spectra were taken using both the G130M/1222 and G160M/1533 grating settings using four FP-POS positions per setting.The data were reduced with the calcos pipeline twice, once using all photons collected and a second time using "night-only" photons to reduce geocoronal emission, which can strongly contaminate O I λ1302 absorption, a key line for our metallicity analysis.A customized set of velocity alignment and co-addition steps was then applied to the data following the procedure outlined in Wakker et al. (2015).
We normalized the spectra around each line of interest using polynomial fits to regions of unabsorbed continuum.We then used the VPFIT software package (Carswell & Webb 2014) to identify and fit Voigt profiles to the HVC absorption features using a chi-squared mini-mization technique.To be considered real, the absorption component was required to appear in two or more metal ions.The data were then binned by five pixels and visually inspected.Summary plots of the O I absorption profiles and H I emission profiles used in the metallicity calculations are shown in Figure 3, with their associated fit parameters listed in Table 2. Fit parameters of all metal lines detected in the HVCs are given in Table 3. Plots of the metal-line absorption and their associated fits are shown in Appendix A.
The O I λ1302 HVC absorption towards HE 0027  is partially blended with MW ISM absorption and has a reduced S/N ratio due to the night-only reduction.These effects made it difficult for VPFIT to obtain a fit for the HVC O I absorption.We tested the fit by requiring the O I linewidth (b-value) to match that of either the Si II at 19.6 km s −1 or C II at 32.6 km s −1 .The Si II b-value resulted in a lower chi-squared value and a better visual match to the O I absorption than the C II b-value, so we adopt b=19.6 km s −1 for the O I absorption measurement for the HVC toward HE 0027.
Due to the reduced S/N ratio in the night-only spectrum of Mrk 969, VPFIT was unable to produce a reasonable fit to the O I λ1302 absorption in HVC G133.5-75.6-294.Fortunately, the geocoronal O I emission in the combined (day+night) spectrum of Mrk 969 is centered at positive velocities while HVC G133.5-75.6-294has a central velocity of −294.1 km s −1 (see Appendix A), far enough away in velocity to be uncontaminated by the airglow.As such, in this sightline we were able to use the combined spectrum to determine a fit to the HVC O I absorption, which is significant at the 3.3σ level.a The allowed velocity range of gas corotating with the MW disk, calculated using methods described in Wakker (1991).
c These fits used a fixed b-value to match that of the Si II for HE0027 and C II Mrk 969 (see Table 3).
d Geocoronal O I emission does not contaminate the velocity range of the HVC towards Mrk 969.Therefore, this fit was made using the combined (day+night) spectrum.All other O I measurements use night-only spectra.

H I Observations
As part of a joint HST-GBT program (GBT project ID HST270191), we obtained frequency-switched singlepointing L-band GBT data for each sight line over a 9. ′ 1 beam.We calibrated the data using standard techniques and corrected for stray-radiation using the methods described in Boothroyd et al. (2011).We fit third to seventh order polynomials to the emission-free channels to remove residual instrumental baselines.We then smoothed the resulting spectra to a channel width of 1 km s −1 .
We fit Gaussians to the H I emission at the velocity of each HVC using custom Python scripts and calculated the column densities (N HI ) of each HVC using the standard relation: where T B (v) is the velocity-dependent brightness temperature in Kelvin, integrated over the emission velocity range in km s −1 .Using the Gaussian fits, we calculate where h is the height of the fit and FWHM is the full width at half maximum.We then calculate the b-value using FWHM=1.665b.The results of the Gaussian fits to the H I emission can be seen in Table 2.A plot showing the H I spectra for each sight line is given in Figure 3.Of the low-ion absorption lines covered, O I λ1302 is considered the most useful for interstellar metallicity measurements, because oxygen has low levels of dust depletion (Jensen et al. 2005;Jenkins 2009) and small ionization corrections (ICs) at large H I column densities (log N HI ≳ 18.5; Bordoloi et al. 2017).Therefore, we use O I λ1302 for all our HVC metallicity measurements.

Cloudy Models
We calculate the ICs by running a customized suite of Cloudy photoionization models (Ferland et al. 2017), which simulate the ionization conditions in the HVCs.The models assume a plane-parallel geometry with uniform gas density.The slab is illuminated by the position-dependent Galactic radiation field presented in Fox et al. (2014), which is based on Bland-Hawthorn & Maloney (1999) and Fox et al. (2005).We also include the cosmic-ray background (Indriolo et al. 2007) and extragalactic background radiation (Khaire & Srianand 2019).Since we do not know the distance to the HVCs, we scale the flux of hydrogen-ionizing photons, Φ(H), to a range of distances from 5 − 150 kpc, specifically 5, 10, 20, 50, 75, 100, and 150 kpc.Φ(H) is calculated based on the Galactic coordinates and distance of the cloud using a custom Python script1 that takes the three-dimensional radiation field from Bland-Hawthorn et al. (2019) and Antwi-Danso et al. (2020).We use the range 5 − 150 kpc to cover nearby clouds and HVCs that are at larger distance estimates of the Magellanic Stream (e.g.Besla et al. 2012).We also run a Cloudy model for each HVC without a Galactic radiation field (i.e., UV background only) to represent an extragalactic cloud.These extragalactic models correspond to our sample of HVCs being at distances ≳150-200 kpc (depending on the cloud's l and b).
For each distance, we run Cloudy with a grid of hydrogen number densities log n H ranging from −3.0 to 0.0 in steps of 0.2 dex, where n H is in cm −3 .To constrain n H , we plot the modeled Si III/Si II ratios against n H and compare them to the measured Si III/Si II value and its ±1σ errors.For more details on the IC error handling, see Appendix B. An example of the modeled Si III/Si II vs. log n H and the modeled ICs vs. log n H is shown in Figure 4, which models the HVC toward CTS 47.We then rerun the Cloudy models with the exact interpolated n H values that match the measured Si III/Si II and the associated ±1σ errors.The ICs are then calculated for each distance and added to the abundance of that ion.In Figure 5 we show an example of the IC dependence on distance, again using the example of the cloud toward CTS 47. Importantly, we find that the modeled ICs do not depend strongly on distance, with the exception of aluminum, whose IC becomes slightly less negative with increasing distance.

Abundances and Depletion
We calculate the gas-phase abundance of element X as: where X i is the observed ion, N X i and N H I are the ionic and atomic hydrogen column densities, respectively, and log (X/H) ⊙ is the elemental solar abundance taken from Asplund et al. (2009).When calculating the oxygenbased metallicity, we add a beam-smearing error to the hydrogen columns of 0.11 dex.This 1σ error is determined by comparing low-and high-resolution H I maps of HVCs, as discussed in detail in Appendix C. We present the HVC oxygen-based metallicity measurements in Table 4.We plot the oxygen, sulfur, carbon, silicon, iron, aluminum, and nitrogen, gas-phase abundances measured in each sight line at each distance in Figure 6 (values are listed in Appendix B).The HVCs each have subsolar gas-phase abundances of each element (uncorrected for dust) and follow a fairly similar abundance pattern, with an average difference between the measured HVC abundances of ∼1.0 dex.The two HVCs showing higher oxygen-based metallicity (the clouds toward HE 0027 and IRAS 0459) consistently have higher abundance measurements than the rest of the HVCs in all elements detected.For oxygen, the HVCs toward HE 0027  and IRAS 0459 have nearly the same abundance, while for most other elements the HE 0027 cloud has a higher abundance than the IRAS 0459 cloud, which could be an indication that the HE 0027 cloud has less dust depletion than the IRAS 0459 cloud in those elements.
In Figure 7 we have separated the abundances into two plots: low-and high-metallicity clouds, and have compared our HVC sample abundance measurements to those of the Magellanic Stream (MS; Richter et al. 2013;Fox et al. 2013).The high-metallicity HVCs closely follow the abundance patterns of the MS.The low-metallicity clouds (towards CTS 47, Mrk 969, and UVQS J0110) only partially fall within the MS abundance pattern.Both their carbon (with the exception of CTS 47) and silicon gas abundances are outside the bounds of the MS abundances.
For each distance, we calculated the dust depletion of each element relative to oxygen, since oxygen is taken to be undepleted onto dust grains: The dust depletion values for each HVC for each assumed distance are listed in Appendix D and plotted in Figure 8.The HVCs all show a similar depletion pattern, with significant depletion seen in carbon, silicon, and aluminum relative to oxygen.Such depletions are usually taken as evidence for dust, but they could also indicate an intrinsically non-solar abundance pattern, as the HVCs may arise outside of the Galactic ISM and therefore have a very different nucleosynthetic history than standard ISM gas.
Only the HE 0027 HVC has an iron measurement (showing very little depletion in iron), while the four other HVCs only show upper limits on iron.The HVCs do not follow a clear pattern of depletion depending on metallicity.The HVCs generally have similar depletion levels to the MS in carbon, silicon, and aluminum (using the MS depletion patterns from Richter et al. 2013;Fox et al. 2013).However, the match is not exact; for example, IRAS 0459 and HE 0027 have MS-like metallicities, but do not match the MS depletions of silicon and iron, respectively.

DISCUSSION
The origins of compact and small HVCs are poorly understood, and are likely to be diverse.They could be nearby clouds in the Galactic fountain, fragments of the Magellanic System, dwarf galaxies with low star formation rates in the Local Group, minihalos, or clouds in the intergalactic medium (Oort 1966;Giovanelli 1978;Bregman 1980;Braun & Burton 1999;Blitz et al. 1999;Brüns et al. 2001;Sternberg et al. 2002;de Heij et al. 2002c;Westmeier et al. 2005;Pisano et al. 2004Pisano et al. , 2007;;Putman 2006;Putman et al. 2012).
Metallicities provide key information to constrain the origin of these clouds, because different origins will lead to different metallicities.Disk clouds and fountain clouds are expected to be metal-enriched, with ∼0.50 solar to super-solar metallicity, whereas halo clouds, such as the well-studied HVC Complex C, have lower metallicities of 0.10 − 0.30 Z ⊙ (Wakker et al. 1999;Gibson et al. 2000;Richter et al. 2001;Tripp et al. 2003;Fox et al. 2023).Magellanic Stream clouds typically have metallicities of ≈ 0.10 Z ⊙ (Fox et al. 2010(Fox et al. , 2013)), but can reach ≈ 0.50 Z ⊙ in certain directions (Richter et al. 2013).Local Group dwarf irregular galaxies have metallicities of ∼0.03 − 0.30 Z ⊙ (Kunth & Östlin 2000).The low-redshift intergalactic medium (IGM) has been modeled and measured to have a metallicity of ∼0.10 Z ⊙ (Ferrara et al. 2000;Tripp et al. 2002;Shull et al. 2003).
Our new measurements show that all five small HVCs in our sample have a low metallicity of ≤ 0.17 Figure 6.Gas-phase abundances measured in each HVC; measurements for low-metallicity and high-metallicity sight lines are denoted by circles and diamonds, respectively, with error bars.Upper limits are indicated by an arrow.The measurements are corrected for ionization using Cloudy models at each distance (5-150 kpc) and one with no MW UV radiation field ("Extragalactic").The gradient in colors represent different distances for the HVCs with darker colors representing smaller distances.The extragalactic model is denoted by a symbol with a dark outline.The grey-dashed horizontal line indicates solar abundance.For each sight line, we connect the various elements with a solid line as a visual aid to show the abundance pattern.
and that three of the five have very low metallicities of ≤ 0.04 Z ⊙ (see Table 4).All of these metallicities are oxygen-based, and so are robust against ionization and dust corrections.We have also performed a thorough analysis of beam-smearing in HVCs and have included the results in our metallicity errors (see Appendix C).These low metallicities indicate that the small HVCs are not ejected MW disk gas and instead originate further away, either in the Galactic halo, the Magellanic System, or the Local Group.
All five small HVCs are projected close to the Magellanic System (see Figure 1), so one simple explanation is that they are all fragments of the Magellanic System that have been detached by tidal or hydrodynamic interactions.This explanation has been favored for other small HVCs in the past, for example the Compact HVC toward the QSO Ton S210 (Sembach et al. 2002;Kumari et al. 2015), a sightline that passes close to the Magellanic System.Westmeier & Koribalski (2008) (Lehner 2002) and [O/H] MS = −1.00±0.09(Fox et al. 2013), though higher metallicities of up to ∼0.50 Z ⊙ have been measured in the LMC filament of the Stream (Richter et al. 2013).
The HVCs towards HE 0027 and IRAS 0459 have metallicities of 0.16 ± 0.07 and 0.17 ± 0.07 Z ⊙ , respectively, similar to the values seen in the SMC and MS.They also have similar relative abundance patterns to that of the MS (see Section 3.2).The HVC towards HE 0027  appears to be projected between two filament in the MS and is at a similar velocity to the surrounding MS gas.Therefore, its velocity, projected location, and abundance pattern indicate that HVC G348.3-83.8-19 is very likely a MS fragment.The HVC towards IRAS 0459 (HVC G224.0-34.3+135) is projected ∼40 • away from the edge of the LMC, but may still be Magellanic given that the Leading Arm and Stream combined reach a total of ∼210 • in length (Nidever et al. 2010) and show ionized gas absorption up to 30 • away from the H I (Fox et al. 2014).Additionally, there is a group of small clouds centered on a Magellanic longitude of ∼15 • and Magellanic latitude of ∼30 • of similar velocities (∼100 km s −1 ; see Figure 1) that could connect HVC G224.0-34.3+135 to the rest of the Magellanic System.Therefore, we conclude that HVC G224.0-34.3+135towards IRAS 0459 is likely a fragment of the Magellanic System.
The other three HVCs in our sample, toward CTS 47, Mrk 969, and UVQS J0110, have much lower metallicities of 0.02 − 0.04 Z ⊙ , which are significantly lower than those measured in the present-day Magellanic System.The cloud metallicities appear too low to ex-  plain with a Magellanic origin even when accounting for the possibility of abundance variations across the Magellanic Clouds, because no radial abundance gradient is seen in H II-region metallicities in either the LMC or SMC (Toribio San Cipriano et al. 2017), and only shallow stellar abundance gradients are seen (Cioni, M.-R. L. 2009;Feast et al. 2010), although we cannot rule out the possibility that the clouds had a steeper abundance gradient before they were stripped.Even in the presence of strong beam-smearing, only one of the three low-metallicity HVCs (G237.2-41.1+146towards CTS 47) has a metallicity close to that of the MS, with Z + 1σ ≈ 0.10 Z ⊙ (see Appendix C).
For these low-metallicity HVCs to originate in the Magellanic System, they would have to have formed early in its history (several Gyr ago), when its chemical abundances were much lower.SMC metallicity histories modeled by Pagel & Tautvaišienė (1998) and Cignoni et al. (2013) indicate that the SMC had a metallicity similar to that of the HVCs toward Mrk 969  and UVQS J0110 (HVC G146.2-77.6-279)∼10 Gyr ago.Similarly, the HVC towards CTS 47 (HVC G237.2-41.1+146)has a similar metallicity to the SMC 7.5 Gyr ago in the models of Tsujimoto & Bekki (2009) or ∼9 − 14 Gyr ago in the early formation history of the galaxies in the models of Pagel & Tautvaišienė (1998); Cignoni et al. (2013).However, in this scenario the unenriched clouds would need to survive over many Gyr without being destroyed or enriched while the rest of the Magellanic System increased in metallicity.This seems unlikely.
Metal mixing could potentially be invoked as an explanation for the very low metallicities.In a metal-mixing scenario, cool clouds can exchange material with the surrounding hot plasma, resulting in a cloud with a metallicity in-between its initial metallicity and that of the surrounding hot halo (Gritton et al. 2014;Gronke & Oh 2018;Heitsch et al. 2022).However, for a cloud to be mixed down from initial metallicities of 0.10 − 0.50 Z ⊙ to metallicities of 0.02 − 0.04 Z ⊙ , the surrounding plasma (the diluting medium) would itself have to have a metallicity of ≤ 0.02 − 0.04 Z ⊙ .Even in the recently discovered Magellanic Corona, Krishnarao et al. (2022) estimate the metallicity to be [Z/H] ≈ −1 or 0.10 Z ⊙ , higher than the values in HVCs G133.5-75.6-29, G146.2-77.6-279, and G237.2-41.1+146.It is therefore difficult to argue for a Magellanic origin for these three HVCs, even after accounting for metal mixing.
An alternative explanation for the small HVCs is that they trace intergalactic gas in the Local Group, beyond the Magellanic Clouds.However, it is unclear whether IGM clouds would have such high H I column densities.Most IGM gas is expected to be in the warm and hot phase due to the ionizing UV and X-ray background, with only < 1% in a cold neutral form and most of the H I reaching columns of only ∼10 17 cm −2 (Pop-ping & Braun 2011;Lockman et al. 2012;Takeuchi et al. 2014;Kooistra et al. 2017), whereas our sample reaches columns of > 10 18 cm −2 .Additionally, most of the lowredshift IGM is expected to be enriched to at least ∼0.10 Z ⊙ by metal ejection from galaxies (with the addition of small amounts of enrichment from Pop III stars; Ferrara et al. 2000;Tripp et al. 2002;Shull et al. 2003), which is 2 − 5 times higher than the 0.02 − 0.04 Z ⊙ measured in the HVCs G133.5-75.6-29, G146.2-77.6-279, and G237.2-41.1+146.For all these reasons, we conclude that the HVCs are not likely to trace the IGM.
Another potential explanation for the three lowmetallicity HVCs (G133.5-75.6-29, G146.2-77.6-279, and G237.2-41.1+146) is that they are related to dwarf galaxies.Two possible scenarios are that the clouds are ultra-faint dwarf galaxies whose stars are currently undetected, or are removed from dwarf galaxies by tidal interactions or ram-pressure stripping (Braun & Burton 1999;Blitz et al. 1999;de Heij et al. 2002c).Typical H I sizes of dwarf galaxies are ≈2-10 kpc (Hunter et al. 2012), so our small HVCs would have to be at distances of ≈60-300 kpc to be dwarf galaxies, which would place them well within the Local Group, inside the Milky Way's virial radius.Extremely metal-poor galaxies (XMPs; < 0.10 Z ⊙ ) are not uncommon, making up ∼20% of faint blue dwarf galaxies (James et al. 2015(James et al. , 2016)).Several diffuse dwarf galaxies have been discovered with metallicities <0.03 Z ⊙ , including Leo P, AGC 198691, Little Cub, and J0811+4730 (Giovanelli et al. 2013;Skillman et al. 2013;Hirschauer et al. 2016;Hsyu et al. 2017;Izotov et al. 2017).Dwarf galaxies can also show considerable metallicity gradients, so that gas on their outskirts can have very low metallicities (Taibi et al. 2022).They also often display structure in their H I content (de Blok et al. 1996;Begum et al. 2006), which could potentially explain the structure we observe in our HVCs (Figure 2).
The key challenge for the dwarf-galaxy explanation is the lack of stellar counterparts.We visually inspected DSS, WISE, and 2MASS images of the HVC fields to search for faint stellar components, but none were found.We also conducted a nearby galaxy search within 5 ′ of each HVC using the NASA/IPAC Extragalactic Database (NED), and did not find any known galaxies.This points away from a dwarf-galaxy origin, although it is possible that enough time may have passed for the stellar component to be separated from the gas by an appreciable distance, especially in the presence of a massive host galaxy (Blitz & Robishaw 2000;Pearson et al. 2016).To find an optical counterpart in this case, kinematically-similar optical counterparts would need to be identified and models of the three dimensional kine-matics for both nearby stellar components and the HVC would need to be conducted, which is beyond the scope of this paper.
Finally, the three low-metallicity HVCs could be gaseous minihalos.Minihalos are small dark-matter halos with no current star formation that accumulate gas in the early universe and are potentially the building blocks of larger galaxies such as the MW (Rees 1986).Minihalos were first suggested as an explanation for HVCs by Braun & Burton (1999) and Blitz et al. (1999); they would naturally lack a stellar counterpart, which has not been found for the three low-metallicity HVCs in our sample.Minihalos can also maintain low metallicities assuming they virialize at high redshift (Wyithe & Cen 2007;Cen & Riquelme 2008), consistent with the three sight lines that have low-metallicities of ≤0.04 Z ⊙ .If the small HVCs are minihalos, then they must have distances of ≤750 kpc, otherwise they would be underpressured, too large (∼10 kpc), and have no analogs in nearby galaxy groups (Pisano et al. 2004(Pisano et al. , 2007;;Sternberg et al. 2002).Starless HVC analogs are not detected in nearby galaxy groups (Pisano et al. 2004(Pisano et al. , 2007)).Therefore, assuming the Local Group is not unique in containing large starless gas-rich dark matter halos, Pisano et al. (2007) find that HVCs must be within distances of ∼90 kpc.Similarly, Sternberg et al. (2002) favor the minihalos being circumgalactic objects with distances of ≤150 kpc assuming they have similar dark matter distributions to dwarf galaxies.
However, if the HVCs in our sample are minihalos close to the MW, then it is unclear how they would retain such low-metallicities, because cloud/corona interactions should have destroyed or disrupted the clouds.Surviving clouds will mix with the hot plasma and end up with a metallicity between that of the original gas cloud and the hot plasma on the timescale of tens of Myr (Gritton et al. 2014;Gronke & Oh 2018;Fielding & Bryan 2022).Lower estimates of the hot-halo metallicity are ≥0.30Z ⊙ , while higher estimates place it at closer to ≥0.60 Z ⊙ (Miller & Bregman 2015;Miller et al. 2016;Gritton et al. 2014;Henley et al. 2017), both of which would be expected to raise the metallicity of the HVCs well beyond their ≤0.04Z ⊙ measurements given tens of Myr.Therefore, we conclude that the three lowmetallicity HVCs could be minihalos only if they are located outside the MW hot halo or entered it fairly recently.

SUMMARY
We have analyzed HST UV absorption-line spectra and GBT H I emission spectra of five small HVCs chosen from GASS in the southern Galactic sky (McClure-Griffiths et al. 2009;Moss et al. 2013Moss et al. , 2017)).The clouds cover between 1 and 2.5 • on the sky.Each of the five HVCs has a UV-bright AGN projected behind their observed H I emission within 1 • of the peak GASS emission.We used these five AGN sight lines together with Cloudy photoionization modeling to determine the ionization-corrected gas-phase abundances of each HVC.Our results are as follows: 1.All five of the HVCs have oxygen-based metallicities <0.17 Z ⊙ .These low metallicities preclude a MW disk origin for these HVCs.
2. The HVCs toward HE 0027  and IRAS 0459 (HVC G224.0-34.3+135)have metallicities of 0.16-0.17Z ⊙ , similar to the SMC and the Magellanic Stream.Given that their metallicities, velocities, and projected locations are all consistent with the Magellanic Stream, these two HVCs are likely fragments of the Magellanic System.

The
HVCs toward CTS 47 (HVC G237.2-41.1+146),Mrk 969 (HVC G133.5-75.6-294),and UVQS J0110 (HVC G146.2-77.6-279)all have very low metallicities of ≤0.04 Z ⊙ .Despite their similar velocities and nearby projected locations to the Magellanic System, we conclude they are not likely part of the Magellanic System since their metallicity is too low.A plausible alternative explanation for these clouds is that they are stripped from dwarf galaxies by tidal interactions or ram-pressure stripping.Alternatively, the low-metallicity HVCs could be gaseous minihalos without star formation, explaining both the low-metallicity of the HVCs and lack of a stellar counterpart.If they are minihalos, then in order to maintain a low metallicity, they would likely reside outside of the MW hot halo.
These metallicity results indicate that two of our HVCs are likely to be fragments of the Magellanic System, while the three low-metallicity HVCs could be extragalactic sources such as starless minihalos or gaseous debris stripped by tidal interactions or ram pressure.Deep optical or infrared observations towards the three low-metallicity HVCs are needed to search for optical counterparts and further narrow the possible explanations for their gaseous properties.Our study concludes that small HVCs, as a class, have a variety of origins, from Magellanic System fragments to objects in the Local Group.
The full Cloudy photoionization model results are shown in Table B1, where the errors on each quantity are based on the Si III/Si II measurement errors.Occasionally, Cloudy was unable to produce a model for the density inferred by the ±1σ Si III/Si II ratio (typically the +1σ value for models with low Φ or higher distances).In these cases, the 1σ log n H was assumed to be symmetric about the measured Si III/Si II ratio.
For the HVC toward HE 0027, the modeled iron ±1σ IC errors were both larger than the measured IC for distances ≤ 75 kpc.In these cases the log n H grid resulted in a local minimum for iron IC near the matching Si III/Si II measurements.For these cases, we chose the higher of the two errors on IC and assumed the error was symmetric about the IC from derived from the measured Si III/Si II ratio.
C. BEAM SMEARING The 9 ′ GBT 21 cm beam probes a much larger region of gas than the HST/COS pencil beam.As such, the median H I column density measured over the GBT beam is likely not representative of the true H I column density along the COS sightline.Here we use the high-and low-resolution H I data (9 ′ and 2 ′ ) of three HVCs presented in Faridani et al. (2014) to estimate the error in our H I column measurements.While this technique does not give us an exact measurement of the beam smearing between a 9 ′ and pencil-beam, it provides a quantitative measurement of the expected beam-smearing effects as the resolution of the H I data decreases.Faridani et al. (2014) present H I data for 3 Compact HVCs (CHVC 070+51-150, CHVC 108-21-390, and CHVC 162+03-186) from the Effelsberg telescope and the Westerbork Synthesis Radio Telescope (WSRT).The Effelsberg 21cm beam is approximately the same size as the GBT H I beam at 9 ′ , while the WSRT provides H I data with a beamsize of ∼2 ′ .For each CHVC Faridani et al. (2014) also feather the low-and high-resolution data together to capture the low-density components measured with the Effelsberg telescope and the high-resolution, high-density emission measured with the WSRT (using the methods described in Faridani et al. 2018).The resulting feathered data have the same synthesized beamsize as the WSRT data.
For each of the CHVCs in the Faridani et al. (2014) sample, we measured the H I column density in both the feathered data cube (2 ′ ) and the Effelsberg data cube (9 ′ ).We use the feathered data since it is the most accurate high-resolution representation of the total H I associated with the CHVCs and we compare to the Effelsberg data since those have very similar resolution to our own GBT observations.We used CARTA (Wang et al. 2020) to make zeroth moment maps of each feathered data cube and placed a grid of ellipse regions (sized to each individual data cube's synthesized beam) over the emission associated with the CHVC (out to ∼2-4σ).The zeroth moment maps and the respective locations of the ellipses can be seen in Figure C2.We then created a grid of circles with the same central locations as those in the feathered data cubes, but with a diameter of 9 ′ for the Effelsberg data cube.Each of the zeroth-moment feathered maps have large noise fluctuations around the edges; therefore, we removed any regions where the 9 ′ circle encompassed a significant portion of this noise.
For each ellipse or circle, we obtained an H I spectrum from the respective data cubes (∼2 ′ ellipses for the feathered data cube and 9 ′ circles for the Effelsberg data cube) and fit a Gaussian(s) to the spectrum.If a spectrum had more than one Gaussian component in either data cube, then those components were checked against the components measured in the other data cube from the region of the same central location.Only components that covered similar velocity ranges in both data cubes and could reasonably be assumed to be part of the same cloud were kept in the analysis.For each ellipse or circle, we added all Gaussian components.Then, for each circle or ellipse with the same central location, we compared the ∼2 ′ feathered to the 9 ′ Effelsberg total H I column measurements.The H I column densities were calculated using Equation 1 and the equation for relating brightness temperature to flux (S):  where ∆α and ∆δ are the FWHM values along the beam's major and minor axes in arcsec.In Figure C2 we plotted the logarithmic difference in the H I columns measured at the two resolutions: log N (H I) 2 ′ -log N (H I) 9 ′ .Positive differences (green) represent a higher H I measurement in the 2 ′ beam (feathered data cube) and negative difference (pink) represent a higher H I measurement in the 9 ′ beam (Effelsberg data cube).
A histogram showing the logarithmic difference for all of the compared regions in all three CHVCs is shown in Figure C3.We fit a Gaussian function to this distribution and found a 1σ dispersion of 0.11 dex.This number is used throughout the paper as the systematic beam-smearing error on our H I measurements.This value is close to 0.15 dex, which is often considered a reasonable estimate on HVC beam-smearing errors (Fox et al. 2018;Ashley et al. 2022).
In the case of our low-metallicity measurements, the highest difference between the 2 ′ and 9 ′ data cubes of 0.54 dex increases the error on the linear metallicity measurement to: 0.044± 0.055 0.044 Z ⊙ , 0.021± 0.026 0.021 Z ⊙ , and 0.019± 0.024 0.019 Z ⊙ for CTS 47, Mrk 969, and UVQS J0110, respectively (see Table 4).Therefore, even in the event that strong beamsmearing causes all our sample HVCs to have an overestimated H I column density (and therefore an underestimated metallicity), the metallicities of the CTS 47, Mrk 969, and UVQS J0110 HVCs are still ≲0.10Z ⊙ and both Mrk 969 and UVQS J0110 retain metallicities <0.05 Z ⊙ .These low metallicities are therefore robust against beam-smearing effects.
We also plot the logarithmic difference between the 2 ′ and 9 ′ H I maps against the 2 ′ H I columns (which provides the most reliable H I measurements) in Figure C4.We see that at low N (H I), the 9 ′ data tends to overestimate the H I column, whereas at high N (H I), the 9 ′ data tends to underestimate the H I column density.As such, when using the larger 9 ′ beam of the GBT, sight lines that pass through the low N (H I) regions of the CHVC are more likely to underestimate the metallicity and sight lines that pass through in the high N (H I) regions of the CHVC are more likely to overestimate the metallicity.While there is significant scatter in this relationship, each of the three background AGNs that result in a low-metallicity measurement (CTS 47, Mrk 969, and UVQS J0110) pass through a central, higher-density region in their respective HVC, providing us with confidence that these three HVCs have genuinely low metallicities.

D. DUST DEPLETION
We list the full dust depletions for each measured ion and assumed distance in Table D2.Note-For each HVC and for each distance modeled, this table reports the oxygen ionization correction, the corrected oxygen abundance, the line-of-sight cloud depth, and the thermal pressure.
Figure 1.Close-up H I maps of the clouds are shown in Figure 2.These clouds are not all strictly considered as CHVCs based on the original definition of Braun &

Figure 1 .
Figure 1.H I 21 cm emission maps of HVCs in Galactic and Magellanic coordinates from HI4PI data HI4PI Collaboration et al. (2016); Westmeier (2017), showing the relative location of the five HVCs (and their background AGN) in this work and their position with respect to the Magellanic System.The AGN are shown as circles.The colors of the AGN circles and the H I emission represent the LSR velocities of the high-velocity absorption/emission.

Figure 3 .
Figure 3. HST/COS O I absorption-line spectra and GBT H I emission-line spectra used for the HVC metallicity measurements.The green lines show Voigt fits to the UV absorption and Gaussian fits to the HVC H I emission.The pink vertical dotted lines represent the velocity centroids of the H I emission and the O I absorption.

Figure 5 .
Figure 5. Ionization corrections (ICs) from Cloudy models over all tested distances for the HVC toward CTS 47. Solid lines represent the ICs and dotted lines represent their associated errors.Filled circles at 150 kpc represent ICs and their errors (vertical lines) for "extragalactic" models without a MW UV field.
and Nidever et al. (2010) suggest that the HVCs towards Mrk 969 and UVQS J0110 are likely part of the MS based on their H I position and velocities alone.For comparison, the SMC and LMC have current-day metallicities 0.22 Z ⊙ and 0.46 Z ⊙ ([O/H] SMC = −0.66 ± 0.10 and [O/H] LMC = −0.34± 0.06; Russell & Dopita 1992; Asplund et al. 2009).The Magellanic Bridge and large parts of the Magellanic Stream have lower metallicities of 0.11 and 0.10 Z ⊙ , respectively, based on the measurements [O/H] MB = −0.96± 0.13 0.11

Figure 7 .
Figure 7. Gas-phase abundances measured for the high-metallicity (top panel) and low-metallicity (bottom panel) HVCs.Comparison data on the Magellanic Stream are shown as grey shaded regions with arrows indicating limits (from the RBS 144 and Fairall 9 sightlines; dark grey and light grey, respectively; Fox et al. 2013; Richter et al. 2013).

Figure 8 .
Figure 8. Depletion relative to oxygen and sulfur in each HVC, compared to the Magellanic Stream depletion pattern.All symbols are the same as in Figure 7.

Figure A1 .
Figure A1.UV absorption-line spectra for metal ions and GBT H I emission spectra for each AGN direction in our sample.The Voigt-profiles fits to each ion and the Gaussian fit to the HVC H I emission are shown in green.The vertical dotted lines represent the velocity centroid of the O I absorption used for metallicity measurements.The H I emission that peaks near ∼250 km s −1 in IRAS 0459 is an artifact in the GBT data and not real.Mrk 969's O I absorption shows all photons.

Figure C3 .
Figure C3.The distribution of the difference in H I column density between the feathered and low-resolution H I maps.Negative numbers represent a larger H I column in the 9 ′ data and positive numbers represent a larger H I column in the 2 ′ data.

Figure C4 .
Figure C4.The relationship between the H I column density measured in the 2 ′ feathered datacube and the difference between the H I column densities measured in the 9 ′ data cube and the 2 ′ feathered data cube.Negative numbers on the x-axis represent a larger H I column in the 9 ′ data and positive numbers represent a larger H I column in the 2 ′ data.

Table 1 .
Basic Sight Line & HVC Information

Table 3 .
Voigt Profile Fits to HVC Absorption 0 (km s −1 ) b (km s −1 ) log N (N in cm −2 ) a 3σ upper limits are derived from the r.m.s.noise measured over the 3σ velocity range on each side of the O I velocity.b The b-value was held constant to obtain a better fit to the absorption line.For the HVC absorption toward HE 0027, O I was held to match that of Si II.For the HVC toward IRAS 0459, bN I was held to match b(O I).For the HVC toward Mrk 969, b(O I) was held to match b(C II).For the HVC toward UVQS J0110, b(Fe II) was derived from Fe II λ1608 alone.

Table 4 .
(Asplund et al. 2009ctions and Metallicity Measurements Gas-phase ion abundance, [X i /H]=[log N (X i )-log N (H I)]-log (X/H)⊙.We use a solar oxygen abundance of 10 −3.31(Asplund et al. 2009).Errors are added in quadrature and include a 0.11 dex error estimate for beam smearing.See Appendix C for details on the calculation of this error.bRange of IC values calculated for a range of distances that the HVCs might plausibly cover.cMetallicities are presented in both logarithmic and linear forms.

Table B1 .
Results from Cloudy Photoionization Models to Small HVCs