Mapping the Supernovae Driven Winds of the Large Magellanic Cloud in Hα Emission I

We surveyed the faint Hα (λ_Hα = 6562.8 Å) emission across the face of the LMC's disk and in the region that surrounds it using the WHAM telescope over the +50 ≤ v_LSR ≤ +250 km s⁻¹ velocity range. ¹¹ Equipped with a Fabry–Pérot spectrometer, WHAM is roughly an order of magnitude more sensitive than other currently available instruments, enabling us to detect the faint emission at a sensitivity limit of I_Hα ≈ 10 mR ¹² per 30 s exposure. WHAM's sensitivity is achieved due to its high throughput resulting from its 1° beam size. By adjusting the gas pressures between the instruments etalon, we can tune our observations to center on the Hα emission line that is associated with the LMC. More information about the WHAM telescope and its capabilities are outlined in Haffner et al. (2003).

Our WHAM observations have a θ_resolution = 1° angular resolution, which corresponds to a Δd_resolution ≈ 1 kpc spatial resolution at the distance of the LMC. The observations were Nyquist sampled across the face of the galaxy, effectively increasing the integration time per location on the sky and removing gaps between observations with overlapping pointings. Our Hα survey contains more than 6600 individual observations that are positioned both on and off the LMC's H i disk, covering an area that spans from (l, b) = (246.5°, −17.8°) to (315.0°, −46.7°).

We use archival H i 21 cm emission-line data from the Parkes Galactic All-Sky Survey (GASS; ¹³ McClure-Griffiths+2009) to probe the neutral hydrogen gas phase in the LMC at an angular resolution of ${\theta }_{\mathrm{resolution}}=16^{\prime}$, corresponding to an angular area that is roughly 14 × smaller than that of the WHAM beam. This survey spans a velocity range of −400 ≤ v_LSR ≤ +500 km s⁻¹ and has a spectral resolution of Δv_resolution = 0.82 km s⁻¹. The rms brightness temperature noise is T_{B, RMS} = 57 mK, corresponding to a 3σ sensitivity of $\mathrm{log}\left({N}_{{\rm{H}}{\rm\small{I}}}/\,{\mathrm{cm}}^{-2}\right)\,\approx 18.2$ for a typical high-velocity cloud line width of 30 km s⁻¹ (McClure-Griffiths et al. 2009). ¹⁴ In this paper, the survey's 3σ limit is far lower than the practical limit used for our maps and mass calculation.

In addition to the GASS data, we use a combined H i 21 cm emission-line map consisting of combined Australia Telescope Compact Array (ATCA) and Parkes telescope observations (Kim et al. 2003). The velocity range of this LMC H i survey spans +190 ≤ v_LSR ≤ +375 km s⁻¹ and has a Δv_resolution =1.6 km s⁻¹ spectral resolution. Compared to the GASS survey, this combined map has a column density sensitivity of $\mathrm{log}\left({N}_{{\rm{H}}{\rm\small{I}}}/\,{\mathrm{cm}}^{-2}\right)\approx 18.86$ for a Δv = 30 km s⁻¹ wide line, but a much higher spatial resolution at ${\theta }_{\mathrm{resolution}}=1^{\prime}$. This data resolves smaller physical structures than GASS, improving our ability to discern smaller-scale morphological features in the disk.

We also use data from the Leiden/Argentine/Bonn (LAB) Survey to measure extinction along our sightlines due to its similar beam size as WHAM. LAB data has an effective FWHM beam of ${\theta }_{\mathrm{resolution}}=35^{\prime}$ for decl. ≤ −27°. The survey covers a velocity range of −450 ≤ v_LSR ≤ +400 km s⁻¹ with a spectral resolution of Δv_resolution = 1.3 km s⁻¹. The rms noise is T_B,rms = 0.07 K resulting in a 3σ sensitivity of $\mathrm{log}\left({N}_{{\rm{H}}{\rm\small{I}}}/\,{\mathrm{cm}}^{-2}\right)\approx 18.38$.

We utilized the WHAM pipeline that is described in detail in Haffner et al. (2003). During this data processing, pixels warmed by cosmic rays were first removed. The circular interference patterns that result from our Fabry–Pérot spectrometer observations were summed in annuli to produce a linear spectrum that is a function of velocity. These linear spectra span a Δv = 200 km s⁻¹ velocity range and are uniformly binned to Δv_bin = 2 km s⁻¹ intervals. The pipeline normalizes the spectra by exposure time, scales them for the air mass of observations, and applies an intensity correction factor to account for sensitivity degradation of the WHAM instrumentation that occurs over time.

Our observations span +50 ≤ v_GEO ≤ +250 km s⁻¹ in the geocentric (GEO) velocity frame. Over this range, these observations do not overlap with the bright geocoronal Hα line at v_GEO = −2.3 km s⁻¹ and only overlaps with the blue wing of a bright OH line at v_GEO = +272.44 km s⁻¹. Therefore, we were unable to use either of these lines to calibrate our velocities using the method described in Hausen et al. (2002) and Barger et al. (2013). Instead, we used the velocity calibration technique that is described by Barger et al. (2017) and Antwi-Danso et al. (2020) for WHAM observations that do not overlap with bright atmospheric lines at well established transitions. Using this technique, we calibrated our velocity by monitoring the pressure of the SF₆ gas in the WHAM Fabry–Pérot etalons and by further refining the calibration by comparing our observations with an atmospheric template.

Using the linear relationship between the pressure of the Fabry–Pérot etalons and Δλ measured by Tufte (1997), we calculated the velocity offset between the raw and GEO velocity frames. This is essentially the reverse of our tuning process, which enabled us to calibrate the velocity frame to an accuracy of Δv_GEO ≲ 5 km s⁻¹. Because all of our observations were taken at the same tune (i.e., at the same interference order), the relative velocities of the calibrated observations agree within 0.1 km s⁻¹ of each other as described in Barger et al. (2017). We improved our calibration further by aligning our observations with the faint atmospheric lines in the atmospheric template presented by Barger et al. (2013) (see their Figure 3). This enabled our velocity solution to be calibrated to an accuracy of Δv_GEO ≲ 1 km s⁻¹. We then converted from GEO to LSR using a constant offset that accounted for the date, time, and location of each observation.

3.3.1. Atmospheric Template

There are significant faint atmospheric emission features that populate the entire velocity range of our observations. While these lines are abundant, they behave predictably and vary primarily with air mass. Because of this, we modeled these lines using a template created by Barger et al. (2013) (see their Figure 3), which characterizes the atmospheric emission present in our observations. The template was scaled to account for differences in air mass between observations and night-to-night variations due to humidity and temperature.

Brighter lines are more variable and need to be fit individually. This includes a bright OH molecular line at v_GEO = 272.44 km s⁻¹ whose line strength depends on the angle between the Sun and Earth's upper atmosphere. In the direction of the LMC, this bright OH line is overwhelmed by emission from the LMC's disk. To subtract this emission feature in the direction of the LMC, we kept the area and width of the line fit constant so that it matched with off-disk observations that were taken during the same night and within roughly 15 minutes of the on-disk observations. Because this narrow OH line is kinematically unresolved by the WHAM telescope, we fit the line assuming that it has a width of Δv_{OH width} = 1 km s⁻¹ before we convolved it with WHAM's Δv_{WHAM IP} ≈ 12 km s⁻¹ instrument profile.

We fit the background emission with a constant term added to the atmospheric template. The total fit—which includes the bright OH line at v_geo = 272.44 km s⁻¹, faint atmospheric lines, a flat background, and Gaussian modeled astronomical emission—utilized a chi-square minimization with conservative criteria to avoid over-fitting emission features that are not physically realistic. These criteria account for the line width and lower limits for the strength of the astronomical emission, which are dependent on the gas temperature and instrument sensitivity, respectively. An example of the pre- and post-atmospheric corrected spectrum is shown in Figure 1 with a sightline piercing through the LMC. A detailed description of these bright lines and how they were handled, including their origin and the nature of their variability, is outlined in Barger et al. (2013).

Zoom In Zoom Out Reset image size

3.3.2. Removal of Systematic Signatures

Following the removal of the atmospheric contamination with the atmospheric template described above, we discovered systematic spectral signatures in the reduced spectra. The cause of these structured residuals is likely due to very faint, unresolved atmospheric lines that are not described in the atmospheric template or from a slight velocity misalignment between the observed spectra and the atmospheric template causing the signatures during subtraction. These signatures appear at the same GEO rest-frame velocity over narrow, 5 ≲ Δv ≲ 10 km s⁻¹, velocity widths. These signatures are visible when the spectra in our survey are stacked across the same velocity range in the GEO frame, as shown in the top panel of Figure 2. At several velocities in this frame, there are multiple relatively coherent vertical signatures across the spectra. However, the two bright astronomical horizontal structures are associated with the Magellanic Bridge (MB) (+150 ≲ v_GEO ≲ +210 km s⁻¹ and 0 ≤ spectrum channel ≤ 35) and the LMC and its wind (+150 ≲ v_GEO ≲ +300 km s⁻¹ and 100 ≤ spectrum channel ≤ 200).

Zoom In Zoom Out Reset image size

These faint residual atmospheric signatures exist across all observed spectra. To characterize these spectral artifacts, we averaged the spectra together at locations within our map that contain little to no astronomical Hα emission above our sensitivity limit. We then subtract this average off-target spectrum from our observed spectra to effectively remove these signatures. To ensure that the off-target spectra accurately characterized the sky for that region of the map, this procedure was repeated for four Galactic latitudinal subregions. We display stacked spectral image with before (top panel) and after (bottom panel) samples of this reduction process for a subset of our observations in Figure 2.

Overall, the vertical atmospheric features at v_GEO ≈ +150, +180, and +230 km s⁻¹ have been greatly reduced. However, some of this residual atmospheric emission remains, especially at v_GEO ≈ +180 km s⁻¹ for spectrum channels > 350 that correspond to a region on the sky between (l, b) = (265°, −20°) and (285°, −25°). These atmospheric residuals persist in these spectra because the Galactic latitude region had few good Hα faint off locations. Although the presence of this lingering atmospheric emission in our final reduced Hα emission map is not ideal, it does not impact the final results of this paper as we do not use the data in that region of the sky in our mass calculation.

We removed sightlines that are within a 0.55° angular radius of stars that are m_V ≤ 6.0 mag as their stellar continuum contributes nonlinear emission to the Hα spectra. This cutoff removes 689 observations from our sample, corresponding to roughly 12% of our observations. Because WHAM is a remote observatory configured for queue observing and does not require constant monitoring during good weather conditions, we double checked all of the observations that were taken at the optimal charge-coupled device (CCD) temperature for the WHAM camera of T_CCD = −101.2 °C and not during an automated liquid nitrogen fill to reduce noise. We confirmed that the etalon pressures were stable and that the monitored values matched the input values during the set up of the observations at night. We also ensured that all of the used observations were taken when the outside humidity was less than 85%, at a zenith distance less than 75°, and during dark time observations. Throughout this process a total of 291 additional observations (or around 4%) of the remaining sample were removed. In total, there were 980 observations (or 14%) removed from our sample leaving 5931 observations. For sightlines with repeat observations, we averaged the spectra together. Our survey samples a total of 1712 unique sightlines, where each sightline was observed an average of 3.5 times with the locations toward the LMC's H i disk sampled the most. Figure 3 shows various spectra contained in our survey.

As a result of WHAM instrument degradation over time, observations suffered up to a 20% decrease in observed Hα intensity (Smart et al. 2019). The procedure used for determining the WHAM instrument degradation trend with time is outlined in Haffner et al. (2003). However, our intensity correction does not include a night-to-night intensity calibration associated with air-mass variations that are due to variations in atmospheric conditions. This is because there were insufficient calibration observations taken each night during this survey, in part because there are few WHAM calibration targets in the southern sky.

Previous absorption-line spectroscopic studies toward LMC stars (e.g., Howk et al. 2002; Lehner & Howk 2007; Lehner et al. 2009; Barger et al. 2017) indicate that the gas clouds surveyed in this study exist between us and the LMC. Additionally, Barger et al. (2017) found compelling evidence that the gas within Δv_LOS ≈ 100 km s⁻¹ from the H i disk of the LMC along the line of sight is associated with large-scale galactic outflow. Similarly, the depletion patterns observed by Lehner et al. (2009) for the HVC material that is in the same projected region on the sky also indicates that this cloud contains dust. We, therefore, applied two extinction corrections: one for attenuation due to the MW and another for self-extinction of the gas clouds assuming that they have a chemical composition similar to that of the LMC.

To correct for reddening, we used the following relationship from Diplas & Savage (1994) that relates the color excess with the average N_{H I} foreground emission:

$$\begin{eqnarray}&&E(B-V)=\displaystyle \frac{\left\langle {N}_{{\rm{H}}{\rm\small{I}}}\right\rangle }{4.93\times {10}^{21}\,\mathrm{atoms}/({\mathrm{cm}}^{2}\,\mathrm{mag})}.\end{eqnarray} \tag{ 1 }$$

We used the LAB Galactic H i survey (Hartmann & Burton 1997; Kalberla et al. 2005) to calculate the average foreground H i emission. For the MW, we integrated the H i emission over the −100 ≤ v_LSR ≤ +100 km s⁻¹ velocity range and assumed an extinction parameter of R_v = 3.1. Similar to the Barger et al. (2013) study, we also adopt the Cardelli et al. (1989) optical extinction curve. Combined, the total extinction for the MW dust is then given by

$$\begin{eqnarray}&&A({\rm{H}}\alpha )=5.14\times {10}^{-22}\left\langle {{\rm{N}}}_{{\rm{H}}{\rm\small{I}}}\right\rangle \,{\mathrm{cm}}^{-2}\,{\mathrm{atoms}}^{-1}\,\mathrm{mag},\end{eqnarray} \tag{ 2 }$$

where the extinction corrected intensity is then I_Hα,corr = I_Hα,obs e^A(Hα). After correcting for extinction associated with foreground MW material, our observed Hα intensities increase by roughly 10%.

The self-extinction by the CGM is small as this gas is diffuse. Following a similar process to the MW extinction, we can account for the self-extinction of the winds by adopting the extinction parameter for the LMC measured by Gordon et al. (2003) of R_v = 3.41. We used an R_v that is measured for the LMC's disk as the IVC and HVC material likely originated from this galaxy via stellar feedback events. When we integrated the H i emission across the velocity range of our wind, +100 ≤ v_LSR ≤ +225 km s⁻¹, we found that the associated self-extinction correction is much less than 1%, and therefore, we neglected this correction in our mass calculations below.

Zoom In Zoom Out Reset image size

Barger et al. (2016) estimated the mass of the intermediate-velocity outflowing LMC winds to be $\mathrm{log}\left({M}_{\mathrm{ionized}}/{M}_{\odot }\right)\,\gtrsim 7.16$ for the low-ionization species. Because they found that this wind is roughly symmetrical on either side of the LMC's disk, this would correspond to a mass of $\mathrm{log}\left({M}_{\mathrm{ionized}}/{M}_{\odot }\right)\,\gtrsim 6.86$ for only the nearside ionized outflow. However, as that study only sampled the wind along two neighboring sightlines via absorption-line spectroscopy, they had to make assumptions about the spatial extent and morphology of the wind. With our kinematically resolved, nearside Hα emission map of the wind of the LMC, we are able to measure both of these directly.

In contrast with the absorption-line work, our Hα survey allows us to obtain a mass estimate more naturally from the wind's density times its volume, M = ρV. We calculate its mass density using the electron number density as a proxy for the density of protons as they are roughly equal (i.e., n_p ≈ n_e ) and use a reduced mass of μ ≈ 1.4m_H to account for the contribution from helium and metals. Calculating the volume of the wind requires knowing its solid angle Ω, line-of-sight depth L, three -dimensional geometry (see Figure 8), and distance D. We also include a $\cos \left(i\right)$ factor to account for the inclination of the cross-sectional area of the wind relative to our line of sight. The mass is then given as: $M=\mu {n}_{e}{\rm{\Omega }}{D}^{2}L\cos \left(i\right)$. For gas at the distance of the LMC, the mass enclosed within one WHAM beam with Ω = 1° is then:

$$\begin{eqnarray}&&\displaystyle \frac{{M}_{\mathrm{ionized}}}{{M}_{\odot }}=2.1\,\times \,{10}^{4}\cos \left(i\right){\left(\displaystyle \frac{D}{50\mathrm{kpc}}\right)}^{2}\left(\displaystyle \frac{{L}_{{{\rm{H}}}^{+}}}{\mathrm{pc}}\right)\left(\displaystyle \frac{{n}_{e}}{{\mathrm{cm}}^{-3}}\right).\end{eqnarray} \tag{ 4 }$$

We estimate the total mass of the wind by summing the single beam mass across the projected outflow area. For the 1712 Hα spectra that fill our map of the LMC (Figure 5), we define the morphological extent of the LMC's galactic wind to include regions that are within 1° of its H i disk with neutral column densities larger than $\mathrm{log}({N}_{{\rm{H}}{\rm\small{I}}}/\,{\mathrm{cm}}^{-2})\geqslant 19.0$. This region contains 215 WHAM sightlines contained within roughly 50 deg² that are used for our mass estimate. For each of these sightlines, we integrated across the −100 ≤ v_LMCSR ≤ −55 km s⁻¹ velocity range to measure the Hα intensity of this wind and to explore its spatial distribution (see Equation (3)). The strength of the Hα recombination line is directly proportional to the electron density squared along the line-of-sight depth and the electron temperature (T_e ) of the gas as

$$\begin{eqnarray}&&\displaystyle \frac{{I}_{{\rm{H}}\alpha }}{{\rm{R}}}=0.364\,{T}_{4}^{-0.924}\left(\displaystyle \frac{\mathrm{EM}}{\mathrm{pc}\,\,{\mathrm{cm}}^{-6}}\right),\end{eqnarray} \tag{ 5 }$$

where EM is the emission measure (EM ≡ ∫n_e (s)² ds) and T₄ is given in units of 10,000 Kelvin (i. e.,T₄ = T_e /10⁴ K). Measurements of the average EM and associated velocity range are given in Table 1.

Since we cannot measure the line-of-sight depth or the electron density as a function of depth directly, we adopt a few necessary assumptions to estimate the mass of the wind that closely follow the procedures used in prior WHAM Hα studies for evaluating the mass of HVCs (e.g., Hill et al. 2009; Barger et al. 2012, 2017; Smart et al. 2019).

6.1.1. Line-of-sight Depth

The most difficult of these assumptions pertains to the depth and line-of-sight distribution of the wind. Past studies of the HVC component of the LMC wind observed similar kinematics for the neutral and low-ionization species (e.g., Lehner et al. 2009). Moreover, observations of both Hα and H i in outflows of other galaxies (e.g., M82; Lehnert et al. 1999 and Schwartz & Martin 2004) support a multiphase wind that is well mixed at large scales. In our study, due to our large angular resolution at 1°, we are spatially resolving the wind at the kiloparsec scale and are unable to resolve small-scale structure in the wind. We therefore assume that the neutral gas and ionized gas are well mixed at the scales we are probing in our survey such that the ionized hydrogen depth is roughly equal to the neutral depth, i.e., ${L}_{{{\rm{H}}}^{+}}\approx {L}_{{\rm{H}}{\rm\small{I}}}$.

To estimate this depth, we analyze the fiducial simulations of Bustard et al. (2020) for LMC-specific outflows (see Figure 7). These magnetohydrodynamic simulations used the observed LMC star formation history from Harris & Zaritsky (2009) to the seed star cluster particles that would subsequently deposit the thermal, kinetic, and cosmic-ray energy into surrounding grid cells. In these simulations, gas that emerged from the LMC's disk due to stellar driven outflows experienced an external pressure by surrounding coronal gas. Bustard et al. (2020) found that the apparent coronal gas wind pushes against the leading edge of the LMC via ram pressure and that its effects are strong enough to suppress the nearside outflows and alter the shape of the LMCs halo. While this simulation neglects the gravitational influence of the SMC and MW, we expect the depth of the outflow to be primarily influenced by the LMCs gravitational potential and ram-pressure effects.

Zoom In Zoom Out Reset image size

On a global scale, the galactic winds produced in the Bustard et al. (2020) simulations match well kinematically and spatially with the observed LMC outflow. We use the Bustard et al. (2020) results as a guide for constraining the depth of this wind. They find that the 10⁴ K gas in the nearside outflow penetrated to a height of z_wind ≈ 3 kpc below the midplane of the LMC (z_midplane = 0 kpc) at a lookback time of 60 Myr. At present-day, they find that this wind stalls at a height of z_wind ≈ 4 kpc due to ram pressure. The outflows on the farside of the LMC, however, are able to travel much further off the disk as ram-pressure effects are weaker on the trailing side of the galaxy. Accounting for the height of the LMC's H i disk (z_{H I ,disk} ≈ 1.75 kpc), this corresponds to a present-day wind depth of roughly 1 ≤ L_wind ≤ 2 kpc off the galaxy.

We used this depth to calculate an average electron density for the LMC's galactic wind using the measured EM as $\langle {n}_{e}\rangle =\langle \mathrm{EM}{\rangle }^{1/2}{L}_{{{\rm{H}}}^{+}}^{-1/2}$. We then used this density to calculate the outflow's mass with Equation (4). Although our main uncertainty involves the depth of the wind, it is important to note that the mass scales as ${M}_{\mathrm{ionized}}\propto {L}_{{{\rm{H}}}^{+}}^{-1/2}$, resulting in only a modest variance in mass when we consider a range of reasonable depths. Moreover, we assume an electron temperature in the range of 0.8 ≲ T₄ ≲ 1.2, which is where the Hα emission peaks. We further assume that the temperature of the neutral and ionized hydrogen gas are roughly equal allowing us to relate the neutral and ionized hydrogen number densities for a given pressure scenario as P_ionized/n_ionized = P_neutral/n_neutral under ideal gas conditions.

Table 1. Observed Velocities and Emission Measures

Outflow Component	vLMCSR	$\left\langle \mathrm{EM}\right\rangle$ a
	(km s−1)	(10−3 pc cm−6)
IVC	−100 to −55	390
HVC	−175 to −100	205

Note.

^a This is an average emission measure across the corresponding velocity range used to calculate the ionized mass across all sightlines considered to be part of the wind.

Download table as:  ASCIITypeset image

Because the LMC is nearly face-on, we cannot constrain how the morphology of the wind varies with depth as we only see its two- dimensional projection on the sky. Therefore, we acknowledge three separate volume scenarios: cylinder, partial outward flaring cone, and partial tapered cone (see Figure 8). For the cylindrical wind in scenario (a), we simply assume that the radius of the wind is constant and matches the extent of the Hα emission. We include the outward flaring partial cone geometry in scenario (b) as it has been observed for other galaxies (e.g., M82); in this scenario, we set the inner cone radius to the stellar radius (2.15 kpc) and the outer radius to match radius of the Hα emission. The tapered, inverted cone in scenario (c) is the geometry that resulted for the nearside LMC wind from Bustard et al. (2020) simulation when they accounted for ram-pressure effects; in this scenario, we match the radii to match the simulation and the Hα observations. For these three geometries, the nearside wind masses would correspond to $\mathrm{log}\left({M}_{\mathrm{ion}}/{M}_{\odot }\right)=7.36\pm 0.14$ for volume (a), $\mathrm{log}\left({M}_{\mathrm{ion}}/{M}_{\odot }\right)=7.10\pm 0.14$ for volume (b), and a mass of $\mathrm{log}\left({M}_{\mathrm{ion}}/{M}_{\odot }\right)\leqslant 7.09\pm 0.14$ for volume (c). As we cannot observationally determine which of these wind geometries better matches with the LMC's nearside galactic wind, we report the values of the cylindrical scenario in the text as it is the simplest volume that requires the least assumptions. We report the values and ranges for the other two volume scenarios in Table 2. To estimate the total mass of the neutral and ionized gas of this wind, we assume the outflow is symmetric on both the nearside and farside of the LMC's disk and that it has an ionization fraction of ${n}_{{{\rm{H}}}^{+}}/{n}_{{\rm{H}}}\approx 0.75$ (Barger et al. 2016). We found the total IVC mass of the wind to be in the range of $7.70\leqslant \mathrm{log}\left({M}_{\mathrm{total}}/{M}_{\odot }\right)\leqslant 7.85$. This is compared to the previous Barger et al. (2016) estimate of $\mathrm{log}\left({M}_{\mathrm{ionized}}/{M}_{\odot }\right)\gtrsim 7.16$.

Zoom In Zoom Out Reset image size

Table 2. Mass Estimates for the IVC and HVC Material of the LMC Outflow

Outflow Geometry	Mion	Mtotal	${\dot{M}}_{\mathrm{outflow}}$	η
	(M⊙)	(M⊙)	(M⊙ yr−1)
IVC
Cylinder	7.28–7.43	7.70–7.85	0.83–1.18	2.44–3.93
Outward cone	7.02–7.17	7.44–7.59	0.46–0.65	1.35-2.16
Inward cone	7.01–7.16	7.43–7.58	0.45–0.63	1.32–2.11
HVC
Cylinder	6.93–7.05	7.35–7.47	0.37–0.49	1.09–1.63

Note. All values for M_total assume a symmetrical wind on the nearside and farside of the LMC with an ionization fraction of n_{H II} /n_H = 0.75 that includes neutral and ionized gas assuming a reduced mass of μ ≈ 1.4m_H to account for helium and metals. We used the values listed in Table 1 to calculate these outflow masses and rates.

Download table as:  ASCIITypeset image

We calculate the mass-flow rate for this wind by assuming an outflow time of (t_outflow ≈ 60 Myr). This is calculated using information regarding the last period of star formation (∼100 Myr) as well as the time necessary for the wind to penetrate through the surrounding medium and travel approximately 2 kpc off the H i disk. This results in a total IVC mass-flow rate of $0.83\leqslant {\dot{M}}_{\mathrm{outflow}}\leqslant 1.18\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$. The mass-loading factor is also calculated to study the ratio of the mass-flow rate to the star formation rate, $(\eta \equiv {\dot{M}}_{\mathrm{outflow}}/{\dot{M}}_{\star })$. We adopt a star formation rate in the range of $0.3\lesssim {\dot{M}}_{\star }\lesssim 0.34\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$ to agree with the star formation history of the LMC (see Figure 11 of Harris & Zaritsky 2009). This results in a mass-loading factor between 2.44 ≤ η ≤ 3.93. Because the mass-loading factor, $(\eta \equiv {\dot{M}}_{\mathrm{outflow}}/{\dot{M}}_{\star })$, is much greater than unity, this indicates that the current star formation state of the LMC is unsustainable such that the galaxy could become quenched if this state is prolonged and if the ejected gas is able to escape.

Barger et al. (2016) characterized the IVC material with respect to the LMC using UV absorption-line spectroscopy toward an LMC disk star and a background QSO. They found the nearside material to have an estimated mass of $\mathrm{log}\left({M}_{\mathrm{low}\mathrm{ions}}/{M}_{\odot }\right)\gtrsim 7.16$ for low-ionization species on both the nearside and farside of the galaxy. This corresponds to an ionized mass of $\mathrm{log}\left({M}_{\mathrm{ionized}}/{M}_{\odot }\right)\gtrsim 6.9$ for gas traveling up to Δv_LMCSR ≈ −100 km s⁻¹ on the nearside of the LMC. Over the same velocity range, we calculated the ionized hydrogen mass to be $\mathrm{log}\left({M}_{\mathrm{ionized}}/{M}_{\odot }\right)\approx 7.36$ (see Section 6). While our estimate is larger than the Barger et al. (2016) nearside mass, it is important to note the discrepancies between our estimates can be attributed to how each study determined the masses.

In the Barger et al. (2016) study, they assumed (1) the wind has an angular extent similar to the LMC's H i disk (R_{H I} ≈ 3.7 kpc), (2) a covering fraction of f_Ω = 0.7 for low-ionization and f_Ω = 0.9 for high-ionization species, and (3) the average global strength for this wind could be represented by the absorption they observed along their two sightlines. We, however, find that the Hα emission of the wind extends around 1° on average off the H i disk, which means that the outflow radius is roughly 1 kpc larger (i.e., R_Hα ≈ R_{H I} + 1 kpc). Their third assumption may result in a significantly underestimated outflow mass as the two sightlines used in the Barger et al. (2016) study probed a relatively quiescent region of the LMC. We emphasized the effect of their assumptions by taking the average emission measure from the same region as the Barger et al. (2016) sightline—in the opposite quadrant as 30 Doradus—and calculating a corresponding mass for an area similar to theirs. Using 〈EM〉 ≈ 0.2 pc cm⁻⁶, the outflow mass would be $\mathrm{log}\left({M}_{\mathrm{ionized}}/{M}_{\odot }\right)\approx 7.1$, which is nearly half as large as our IVC cylindrical mass estimate and nearly in agreement with the Barger et al. (2016) estimate for the low-ionization species.

Ultimately, the fate of this ejected material remains uncertain. The escape velocity of the LMC is roughly 110 km s⁻¹ (Besla 2015), which corresponds to roughly 100 km s⁻¹ along the line of sight for an inclination of around 25°. Therefore, the majority of the IVC material is not expected to escape. However, the Magellanic System is a crowded environment and tidal interactions between the SMC, and possibly the MW, can assist in the removal of otherwise bound gas (see D'Onghia & Fox 2016 for a review). Some of the material from previous outflow events may have been displaced into the trailing Magellanic Stream. In a kinematic investigation of the H i 21 cm emission of the Magellanic Stream, Nidever et al. (2008) found that one of its two filaments traces back to the 30 Doradus region of the LMC. Richter et al. (2013) measured the chemical composition of this "30-Doradus" filament and found that it has a metallicity that is consistent with an LMC origin. Fox et al. (2013) explored the chemical composition of the other Magellanic Stream filament and found that it has a lower metallicity that is more consistent with an SMC origin, which kinematically traces back to the Magellanic Bridge (Nidever et al. 2008).

Ram-pressure stripping may also play an important role in removing gas from the Magellanic Clouds. The impact that ram pressure has on the CGM strongly depends on the density of the medium that it is traveling through and on the motion of the gas within its surrounding medium. Based off the work of Salem et al. (2015), Bustard et al. (2020), and others (Heckman et al. 2000; Mastropietro et al. 2005; Fujita et al. 2009), the effects of ram pressure on gas is multifaceted and either can promote or suppress the removal of gas from a galaxy depending on the circumstance. This is because ram pressure can work in direct opposition of galactic winds positioned on the leading side of a galaxy, where the surrounding coronal gas will act as a headwind that pushes against the outflowing back toward the galaxy. Therefore, the outflow on the nearside of the LMC galaxy, which is the side leading the LMC's orbit through the MW's halo, will be suppressed. Meanwhile, the farside could experience an enhanced outflow as ram pressure will push the gas away from the galaxy and it could therefore be more massive. With ram-pressure stripping, the simulated wind in the Bustard et al. (2020) study was able to reach a height of more than 1.7 kpc off the LMC's H i disk and had a total ejected mass in the range of $6.7\,\leqslant \mathrm{log}\left({M}_{\mathrm{ejected}}/{M}_{\odot }\right)\leqslant 8.4$. While our IVC mass estimate is within this mass range, Bustard et al. (2020) predicts that a significant portion of this material flows off the farside of the LMC's disk (see their Figure 11), which is the trailing side trails as the LMC traverses the MW's halo. Conversely, the Barger et al. (2016) UV absorption-line study found that the nearside and farside outflows were kinematically symmetric along their explored sightlines, though this might not be representative of the outflow's large-scale structure. A study like this one of the farside—in which the winds are mapped—is needed to determine its physical extent and the impact of ram-pressure stripping.

We observe high-velocity material toward the LMC at velocities greater than 100 km s⁻¹ off the LMC H i disk (+90 ≤ v_LSR ≤ +175 km s⁻¹), detailed in Section 7. This emission is persistent at intermediate to low velocities relative to the LMC and spatially aligns well with the LMC's H i disk (see Figure 6). These observed properties are consistent with an LMC origin in which the gas is expelled from the galaxy by its stellar activity. This is a conclusion that has also been reached by Staveley-Smith et al. (2003) using an H i emission line and by Lehner et al. (2009) using UV absorption lines (also see Lehner & Howk 2007 and Barger et al. 2016). Staveley-Smith et al. (2003) found that the H i column densities of the HVC peaked at locations that align with H i voids within the LMC disk (such as supergiant shells, e.g., LMC 3); they further identified spatial and kinematic H i bridges that linked back to the LMC's disk. Lehner et al. (2009) used 139 stars embedded in the LMC as background targets in an UV absorption-line study to explore the properties of this HVC; they found that the HVC has (i) an LSR velocity gradient in R.A. that follows the LMC's velocity gradient, (ii) dust based on depletion patterns—signifying a galactic origin (see also Smoker et al. 2015), (iii) an oxygen abundance similar to the LMC of $[{\rm{O}}\,{\rm\small{I}}/{\rm{H}}\,{\rm\small{I}}]=-{0.51}_{-0.16}^{+0.12}$, and (iv) a high covering fraction in the direction of the LMC.

However, since the works mentioned above, the origin of this HVC has been strongly debated. This is because Werner & Rauch (2015) found C ii, Si ii, and Si iii absorption consistent with this HVC in the spectra of a background star at a distance of ${d}_{\odot }={9.2}_{-7.2}^{+4.1}\,\,\mathrm{kpc}$. Richter et al. (2015) confirmed the presence of the HVC in the direction of this star, which places the HVC at a distance of d_⊙ < 13.3 kpc, and asserted that the HVC lies too far from the LMC to be consistent with an LMC origin. They reason that it is unlikely that material ejected from the LMC would be able to reach a distance of roughly 40 kpc intact during the few hundred million years it would take to travel at speeds of ∼ 150 km s⁻¹. This is because, during the cloud's journey it would sustain numerous collisions with not only the CGM of the LMC, but also the halo of the MW. Consequently, Richter et al. (2015) argue that this would strip a large amount of the cloud's material and drag forces would slow the cloud. Needless to say, the time required to make this journey (250–400 Myr; Barger et al. 2016) would consequently lead to a transverse displacement and the cloud would no longer be toward the LMC.

In light of the results presented in this paper, and by previous studies, we offer a mutual theory on the distribution and association of material observed between the MW and LMC. We postulate that there are two HVCs with different origins near the LMC on the sky: (1) an HVC is associated with the galactic winds of the LMC and (2) an HVC that is associated with the MW. Strong evidence for this latter HVC was presented by Werner & Rauch (2015) and Richter et al. (2015), who confirmed that there is high-velocity material at (l, b) = (279.9°, −37.1°), 3°–5° away from 30 Doradus, that resides at a distance of d_⊙ < 13.3 kpc. This sightline is on the periphery of the Hα emission that spans across the LMC (see Figure 5). Meanwhile, similarities in kinematics, dust depletion, and oxygen abundances strongly indicate that most of the high-velocity material in the direction of the LMC has an LMC origin. Unless the previous evidence gathered from Staveley-Smith et al. (2003), Lehner et al. (2009), and our study is entirely coincidental, there is likely to be more than one complex toward the LMC. Therefore, we argue that the LMC HVC spans a much larger area of the sky in the direction of the LMC and that there is also a smaller HVC positioned just offset from the LMC on the sky, which is associated with the MW.

Galactic winds and the ability of a galaxy to retain ejected gas are directly related to the galaxy's capacity to form future stars. These winds tend to intensify as stellar activity increases. Moreover, because lower mass galaxies have smaller gravitational potential wells, their gas is more easily ejected out of them. For massive galaxies, the halo and extraplanar gas that surrounds their disks suppresses outflowing material. This results in a general trend in which the mass-loading factor is diminished for massive galaxies and enhanced in low-mass galaxies. In the case of the LMC, with a stellar mass of M_⋆ = 3 × 10⁹ M_⊙ (van der Marel et al. 2009) and total mass M_total = 1.7 × 10¹⁰ M_⊙ (van der Marel & Kallivayalil 2014), as well as an active star formation history (Harris & Zaritsky 2009), it is expected that this galaxy will have a relatively elevated mass-loading factor.

We estimate that the LMC's mass-loading factor ranges from 3.53 ≤ η ≤ 5.56 when including both the IVC and HVC gas in its winds and when assuming cylindrical geometry. Our values are relatively large when compared to other studies for galaxies with similar stellar mass (see Figure 9). However, the Barger et al. (2016), Chisholm et al. (2017), and Leethochawalit et al. (2019) studies all used absorption-line spectroscopy, which spatially probes less of the wind; this results in a more uncertain mass estimate as the physical extent of winds are poorly constrained. In the case of the McQuinn et al. (2018) Hα imaging study, although they were able to measure the extent of the wind, their observations were more than a magnitude less sensitive than ours and were unable to detect the diffuse gas in the wind.

Zoom In Zoom Out Reset image size

The impact of geometry assumptions is most telling when comparing our results with those of the Barger et al. (2016) absorption-line study of the LMC. Although our study shares many of the same assumptions as their study, we were able to map the extent and morphology of the wind, whereas they were forced to make simplistic assumptions for its solid angle. The mass-flow rates from the Barger et al. (2016) and the star formation rates from Harris & Zaritsky (2009) correspond to a mass-loading factor for LMC that is in the range of 0.7 ≤ η ≤ 0.8. When we increase their solid angle to match what we observe in Hα emission, their mass-loading factor becomes 1.3 ≤ η ≤ 1.5. This revised range is in better agreement with our mass-loading factor for conical geometries (outward and inward cone; see Table 2).

In studies that explore a wide range of galaxy masses, the trend that the mass-loading factor decreases with galaxy mass is clear (see Figure 9). Chisholm et al. (2017) found mass-loading factors ranging from 0.2–19.0 for their sample of eight dwarf star-forming galaxies ($6.9\leqslant \mathrm{log}\left({M}_{\star }/{M}_{\odot }\right)\leqslant 10.7$) using UV absorption-line spectroscopy. At the LMC mass, this study predicts a mass-loading factor of around 1.1. Across similar masses, McQuinn et al. (2018) observed their galaxies via Hα imaging and found mass-loading factors ranging from 0.2 ≤ η_ionized ≤ 7.1 for the ionized gas in the wind only. For more massive galaxies ($9.5\leqslant \mathrm{log}\left({M}_{\star }/{M}_{\odot }\right)\leqslant 11.5$), Leethochawalit et al. (2019) found mass-loading factors that range from 0.3–1.0 for their UV absorption-line study. While our largest LMC mass-loading factor estimate is at least twice as large as the Chisholm et al. (2017) and Leethochawalit et al. (2019) trend lines predict, we acknowledge that is in part because of differences in the assumed geometries of the winds. When assuming a conical geometry, a commonly assumed geometry for galactic winds, our mass-loading factor is reduced by nearly 50% and is more consistent with their results (see Table 2).

Still, there is a significant spread in mass-loading factors between studies that requires additional work to reduce. As of right now, no uniform sample exists across low- and high-mass galaxies. Such a consistent sample would go a long way in improving our understanding of the relationship between mass-loading factor and galaxy mass. Further, more imaging or emission-line studies are needed to better constrain the geometry of these winds, though they also need to have a high enough sensitivity to detect the diffuse gas.