SMASHing the LMC: Mapping a Ring-like Stellar Overdensity in the LMC Disk

The SMASH is a Dark Energy Camera (DECam; Flaugher et al. 2015) survey that has completely mapped the main bodies of the Clouds, and probed the Magellanic Periphery and the Leading Arm region by sampling "island" fields over a ∼2400 deg² area of the sky, at an ∼20% filling factor. The goal of the deep photometric (ugriz ∼ 24th mag) survey is to study the stellar structure and star formation history of the Clouds, with particular focus on low-density stellar populations associated with the stellar halos and tidal debris of the Magellanic Clouds. Using old, main-sequence (MS) stars as a tracer, a technique previously adopted by the Outer Limits Survey (Saha et al. 2010), SMASH data should be able to reveal the relics of the formation and past interactions of the MCs down to surface brightnesses equivalent to Σ_g = 35 mag arcsec⁻² over a vast area.

Nidever et al. (2017) described the reduction and calibration of the data as well as the first public data release, which contains ∼700 million measurements of ∼100 million objects in 61 deep and fully calibrated fields from the SMASH survey.¹⁸ The photometric precisions of the final SMASH catalogs are roughly 1.0% (u), 0.7% (g), 0.5% (r), 0.8% (i), and 0.5% (z). The obtained calibration accuracies are 1.3% (u), 1.3% (g), 1.0% (r), 1.2% (i), and 1.3% (z). The median 5σ point-source depths in the ugriz bands are (23.9, 24.8, 24.5, 24.2, 23.5) mag. The astrometric precision is ∼15 mas and accuracy is ∼2 mas with respect to the Gaia DR1 astrometric reference frame (Gaia Collaboration et al. 2016). We refer the readers to Nidever et al. (2017) for further details on the survey, data reduction, and photometry.

In this study, we use 62 fields¹⁹ that cover the LMC main body, and use the red clump (RC) and bright MS stars to study the LMC's structure. In each field catalog, we first select point-like sources using the DAOPHOT morphology parameters "chi" and "sharp": chi < 3 (<5 for Fields 36 and 41 in the center) and $\,| \,$sharp $\,| \,$ < 1 (<2 for Fields 36 and 41). Contamination by unresolved background galaxies and the foreground MW stars inside the LMC main body is negligible, especially within the magnitude range we are interested in.

The DES Data Release 1 (DR 1) was made earlier this year (Abbott et al. 2018), and it is easily accessible through the NOAO Data Lab.²⁰ In addition to the SMASH data, we use DES data near the LMC main body to better constrain the exponential disk by adding coverage of the northern disk up to ρ = 95 from the center (see Figure 1). Beyond ∼95 to the north, the number density of LMC RC stars decreases rapidly and the level of contamination from MW foreground becomes severe. Thus, we only make use of the DES data inside ρ = 95 for modeling the LMC disk.

Zoom In Zoom Out Reset image size

The RC is one of the most prominent features in the LMC color–magnitude diagram (CMD). RC stars are in the core He-burning stage, and have relatively low masses with intermediate ages on average (≲2 Gyr; Castellani et al. 2000) and moderately high metallicity. In the CMD, RC stars occupy a well-defined, narrow region due to their uniform core mass, regardless of their initial mass. This fundamental property results in very limited effective temperatures and luminosities, allowing RC stars to be used to accurately measure distances and extinctions (Girardi 2016, and references therein).

We follow the procedure used in Choi et al. (2018) to select LMC RC stars. Briefly, we carefully define the RC selection box rather than using a simple rectangular region around the RC (see Figure 3 in Choi et al. 2018) to minimize the contaminants in the RC selection. To do that, we define a large initial box around the RC with a g-band magnitude range between 18.5 and 21. The color range varies for each field according to its dust extinction, while the slope is fixed along the extinction vector. The maximum color is set by the contamination by MW foreground stars. The number of unique SMASH LMC RC stars is ∼2.2 million.

The RC selection from the DES data is nearly identical to that used for the SMASH data. One difference is that we select the RC field-by-field from the SMASH data where dust and variations in stellar population govern the RC morphology, while we select the RC from the DES data in four azimuthal slices to have enough RC stars to be clumped in a CMD and to minimize possible contamination at the same time. The RC morphology does not change very much in the outer LMC disk due to rather homogeneous populations and lack of internal dust, but the shift in the magnitude with position angle due to the disk inclination is not negligible. Thus, we slice the DES data azimuthally to limit the shift in the RC magnitude by 0.1 mag based on the predicted magnitude variation map generated with the inclination and line-of-nodes measurements from Choi et al. (2018). This allows us to reduce the contamination from non-RC stars by keeping the RC selection box fairly narrow. The solid arcs in Figure 1 show the predicted magnitude variation with respect to the fiducial magnitude. The LMC line of nodes is the tangent line to the Δ m = 0 mag arc at the LMC center.

The left panel of Figure 1 shows the star count map of the selected RC stars from both the SMASH and DES data. There is a smooth continuation between the two data sets. The orange line outlines the region contiguously covered by the SMASH data. The combined SMASH and DES area enables us to encompass a sufficient portion of the LMC disk to cover an additional ∼4° beyond the ring-like structure detected in the literature, particularly the northern and southern parts. We refer the readers to Choi et al. (2018) for a more detailed discussion on the intrinsic RC properties, our RC selection in CMDs, and the resulting RC star count map. In the right panel, we present the 1D RC stellar density radial profiles for 11 azimuthal bins with a 2° width within the sector denoted in the left panel. This sector covers the largest radial extent (up to 105) in our map. Because the direction of the sector is roughly parallel to the bar minor axis, the contribution of the bar to the radial profile is limited to a small radius (<15) in this sector. All 11 radial profiles are well described as a single exponential disk (red dashed-line), showing an excess of RC stars at around 6° from the LMC center. Majewski et al. (2009) and Nidever et al. (2018) also showed that the LMC density radial profile is best matched by the exponential disk profile out to ∼9° and ∼13°, respectively. This, combined with the sufficient coverage of the LMC disk, gives us confidence that our data coverage will allow reliable characterization of the overdensity.

The top left panel of Figure 2 shows the combined Hess diagram of SMASH LMC bright MS stars. For bright MS, we select stars with g-band magnitudes brighter than 20 mag and −0.8 < g − i < 0.3. The gray dashed polygon denotes our selection box. The number of selected unique bright MS stars is ∼3 million. We do not include the DES data to construct a list of bright MS stars because the SMASH data alone provide sufficient information on the spatial distribution of bright MS stars and there is no significant number of bright MS stars in the DES region.

Zoom In Zoom Out Reset image size

Figure 2 shows the spatial distribution of selected bright MS stars in a range of magnitude bins. The maximum age in each bin increases from ∼100 Myr to ∼1.8 Gyr. While the RC stars show a relatively smooth and extended spatial distribution, the young MS stars are highly structured and mostly confined to the inner disk. The MS stars in brighter magnitude bins (g < 17.5 mag) are distributed in clumpy structures, while fainter ones (17.5 < g < 19 mag) form coherent structures such as the central bar and one prominent spiral arm, indicating hierarchical star formation (e.g., Harris & Zaritsky 2009; Sun et al. 2017). However, the latter structures are only evident when looking at MS stars fainter than ∼18 mag. The lack of the bar structure in the youngest bins agrees with the detailed star formation histories of the LMC bar (Monteagudo et al. 2018). The MS stars fainter than ∼19 mag extend to even larger radii and start resembling the RC spatial distribution. The maps of bright MS stars as a function of their apparent magnitude suggest that the star-forming disk has dramatically shrunk over the last ∼1–2 Gyr from a radius of ∼7° to ∼4°. This is consistent with the outside-in LMC evolution scenario (Meschin et al. 2014).

In this section, we briefly introduce each of the three coordinate systems used in this work following van der Marel & Cioni (2001) and Mackey et al. (2016): angular coordinates (ρ, ϕ), tangent plane coordinates (ξ, η), and the galaxy plane coordinates (x, y, z). For coordinate transformation from the celestial coordinates (R.A. α, decl. δ) and distance D, we adopt the LMC center as the origin with (α₀, δ₀) = (8225, −6950) and D₀ = 49.9 kpc (van der Marel 2001; de Grijs et al. 2014). We adopt the center of the bar as (α, δ) = (7991, −6945) (de Vaucouleurs & Freeman 1972). In addition, we use inclination and line-of-nodes position angle (i, θ) = (2586, 14923) as determined by Choi et al. (2018) using the 3D spatial distribution of the RC.

Angular coordinates (ρ, ϕ) are defined on the celestial sphere around the origin point (α₀, δ₀) as follows:

$$\begin{eqnarray}\begin{array}{rcl}\cos (\rho ) & = & \sin (\delta )\sin ({\delta }_{0})+\cos (\delta )\cos ({\delta }_{0})\cos (\alpha -{\alpha }_{0})\\ \tan (\phi ) & = & \displaystyle \frac{\cos (\delta )\sin ({\delta }_{0})\cos (\alpha -{\alpha }_{0})-\sin (\delta )\cos ({\delta }_{0})}{\cos (\delta )\sin (\alpha -{\alpha }_{0})},\end{array}\end{eqnarray} \tag{ 1 }$$

where ρ is the angular distance of a point (α, δ) from the origin, and ϕ is the position angle measured counterclockwise from the west at constant decl. δ₀.

The (ξ, η) tangent plane coordinates are obtained through the gnomonic projection of the sphere onto a tangent plane:

$$\begin{eqnarray}\begin{array}{rcl}\xi & = & \displaystyle \frac{\cos (\delta )\sin (\alpha -{\alpha }_{0})}{\sin ({\delta }_{0})\sin (\delta )+\cos ({\delta }_{0})\cos (\delta )\cos (\alpha -{\alpha }_{0})}\\ \eta & = & \displaystyle \frac{\cos ({\delta }_{0})\sin (\delta )-\sin ({\delta }_{0})\cos (\delta )\cos (\alpha -{\alpha }_{0})}{\sin ({\delta }_{0})\sin (\delta )+\cos ({\delta }_{0})\cos (\delta )\cos (\alpha -{\alpha }_{0})}.\end{array}\end{eqnarray} \tag{ 2 }$$

The observed and modeled stellar densities are compared in 10' × 10' cells in this tangent plane (see Section 4.3).

A Cartesian coordinate system (x, y, z) is used to represent the LMC disk plane. The (x, y)-galaxy plane is defined as an infinitely thin plane that is inclined with respect to the sky plane by an angle i about the line of nodes with a position angle θ. We compute the expected stellar number density at positions in the disk plane. The coordinates are

$$\begin{eqnarray}\begin{array}{rcl}x & = & D\sin (\rho )\cos (\phi -\theta )\\ y & = & D[\sin (\rho )\sin (\phi -\theta )\cos (i)+\cos (\rho )\sin (i)]-{D}_{0}\sin (i)\\ z & = & D[\sin (\rho )\sin (\phi -\theta )\sin (i)-\cos (\rho )\cos (i)]+{D}_{0}\cos (i).\end{array}\end{eqnarray} \tag{ 3 }$$

In the LMC galaxy plane (i.e., z = 0), the line-of-sight distance (D) is

$$\begin{eqnarray}&&D=\displaystyle \frac{{D}_{0}\cos (i)}{[\cos (i)\cos (\rho )-\sin (i)\sin (\rho )\sin (\phi -\theta )]}.\end{eqnarray} \tag{ 4 }$$

A single, symmetric exponential disk requires four fitting parameters: central stellar number density (μ_0,d), disk scale length (r_d), axis ratio (b/a), and semimajor axis position angle (ψ):

$$\begin{eqnarray}&&{\mu }_{d}(r)={\mu }_{0,d}\,\exp \left(-\displaystyle \frac{r}{{r}_{d}}\right),\end{eqnarray} \tag{ 5 }$$

where r is

$$\begin{eqnarray}&&\begin{array}{l}r{(x,y)}^{2}\\ \,=\,{(x\cos (\psi )-y\sin (\psi ))}^{2}+{\left(\displaystyle \frac{x\sin (\psi )+y\cos (\psi )}{b/a}\right)}^{2}.\end{array}\end{eqnarray} \tag{ 6 }$$

In contrast to an exponential disk, it is non-physical to model a bar as an inclined 2D structure because of its significant z-direction thickness compared to its x,y dimension. Thus, we model the 2D bar as it appears in the tangent plane.

For the 2D elliptical bar, there are seven fitting parameters, bringing the total number of parameters included in the two-component modeling to 11. We adopt the modified Ferrer profile for a bar (e.g., Laurikainen et al. 2007; Peng et al. 2010),

$$\begin{eqnarray}&&{\mu }_{b}(r)={\mu }_{0,b}{\left[1-{\left(\displaystyle \frac{r}{{r}_{\mathrm{out}}}\right)}^{2-\beta }\right]}^{\alpha },\end{eqnarray} \tag{ 7 }$$

where μ_0,b is the bar central number density, r_out is the end of the bar profile, α determines the sharpness of the truncation at r_out, and β determines the inner slope of the profile.

To account for the boxiness/diskiness, we apply a generalized ellipse (Athanassoula et al. 1990) as follows:

$$\begin{eqnarray}\begin{array}{rcl}r(x,y) & = & ({| x\cos (\psi )-y\sin (\psi )| }^{{C}_{0}}\\ & & {\left.+{\left|\displaystyle \frac{x\sin (\psi )+y\cos (\psi )}{b/a}\right|}^{{C}_{0}}\right)}^{\tfrac{1}{{C}_{0}}}.\end{array}\end{eqnarray} \tag{ 8 }$$

The C₀ parameter determines the shape of the bar: C₀ = 2—purely elliptical; C₀ > 2—boxy; C₀ < 2—disky. The bar ellipticity can be determined from the b/a ratio,  = 1 − (b/a).

The final model prediction in each cell (ij) is the sum of the projected disk component in the tangent plane and the bar component:

$$\begin{eqnarray}\begin{array}{rcl}{\mu }_{m,{ij}} & = & {\mu }_{d,{ij}}+{\mu }_{b,{ij}}\\ & = & \displaystyle \frac{\cos {(\rho )}^{3}}{\cos (\rho \sin (\phi -\theta )+i)}{\left(\displaystyle \frac{D}{{D}_{0}}\right)}^{2}{\mu }_{d,{xy}}+{\mu }_{b,{ij}}.\end{array}\end{eqnarray} \tag{ 9 }$$

An extra term in front of μ_d,xy, which is the exponential disk central number density in the galaxy plane, takes care of the factors from the gnomic projection of the predicted number of stars in a given area on the galaxy plane at a distance D, as well as a distance gradient across the given cell in the tangent plane.

We use only the RC stars, not the bright MS stars, as tracers of the LMC exponential disk and bar, as they sample a population that should be dynamically well mixed. We decompose the LMC into a disk and bar using a Bayesian technique with the Markov chain Monte Carlo (MCMC) sampler emcee²¹ (Foreman-Mackey et al. 2013). With the Bayesian approach, we can extract the posterior probability distribution functions (pPDFs) of each of the model parameters, as well as the covariance between parameters. We can also incorporate prior knowledge about the model parameters. We use flat priors within physically meaningful ranges and zero outside those ranges (see Table 1).

Table 1.  LMC Disk Parameters

Parameter	Description	Range for Flat Prior	Results
μ0,d [Nstars/kpc2]	Disk central number density	$0\mbox{--}\infty$	${2800.89}_{-7.38}^{+7.35}$
rd [kpc]	Disk scale length	$0\mbox{--}\infty$	${1.667}_{-0.002}^{+0.002}$
(b/a)disk	Disk axis ratio	0–1	${0.836}_{-0.001}^{+0.001}$
PAdisk [degree]	Disk major axis position angle	0–2π	${227.24}_{-0.188}^{+0.178}$
μ0,b [Nstars/''2]	Bar central number density	$0\mbox{--}\infty$	${2474.26}_{-14.73}^{+14.21}$
rout [degree]	Bar length	0–3.5	${3.499}_{-0.000}^{+0.000}$
α	Sharpness of the truncation of bar profile	0–10.5	${1.371}_{-0.012}^{+0.012}$
β	Inner slope of bar profile	0–2	${0.427}_{-0.020}^{+0.021}$
C0	Bar shape parameter	1.5–3.0	${2.999}_{-0.000}^{+0.000}$
(b/a)bar	Bar axis ratio	0–1	${0.446}_{-0.001}^{+0.001}$
PAbar [degree]	Bar position angle	0–2π	${154.18}_{-0.11}^{+0.11}$

Download table as:  ASCIITypeset image

Specifically, we impose a conservative prior on bar size, motivated by its apparent morphology and the model prediction (<35; van der Marel 2001; Subramaniam 2004; Zaritsky 2004; Besla et al. 2012). The bar size can be inferred from the disk geometry as well. In the inner disk where the bar dominates the disk morphology, the warped bar induces a larger inclination, and twists the position angle of the line of nodes with respect to the rest of the LMC disk. In particular, the position angle changes rapidly at 2° < ρ < 35 and stays flat beyond that (see Figure 13 in Choi et al. 2018), indicating the maximum length of the bar in the RC star count map is unlikely to be larger than 35. van der Marel (2001) reported very similar behavior in their RGB star count map. Thus, we set the maximum bar size to be 35. Furthermore, without setting this limit, the bar model always tries to fit substructures in the inner disk (e.g., sub-dominant spiral arm and even a part of the ring-like overdensity) as much as possible, which is not physically meaningful.

We also impose a uniform prior on the boxiness (1.5 < C₀ < 3) to consider the range that roughly describes the stellar density contours of the LMC bar region (see the black contours in the left panel of Figure 1). The density contours of the bar region do not suggest significant deviation from a pure ellipse (i.e., C₀ = 2); the shape of the bar is neither disky nor very boxy. Gadotti (2011) analyzed ∼300 barred galaxies in the local universe and found C₀ ranges between 1.5 and 3.5 with a peak at 3. Kim et al. (2015) characterized the shape of bars in 144 face-on barred galaxies, and found the maximum C₀ value of ∼3.5 in their sample. Our adopted prior range for C₀ is well within the range of values found for external barred galaxies.

We conduct a rigorous test on the effect of our choice of the prior ranges by modeling the disk with many different combinations of prior ranges on the bar size (3° < r_out,max < 8°) and its boxiness (3 < C_0,max < 8). The existence of the ring-like overdensity, its overall shape, and the maximum amplitude turn out to be robust against the variations in the bar parameter priors. The position angles of both the exponential disk and bar, and the bar ellipticity, are also insensitive to the prior ranges. The disk axis ratio and the scale length do not vary significantly either. If we allow a more generous range for C₀, the resulting bar always becomes very boxy (C₀ ≳ 3.5) to make up for non-bar structures, such as a sub-dominant arm emanating from the east end of the bar and a part of the ring-like overdensity regardless of the allowed range of the bar size. When keeping the same range for the bar size, the boxiness is the only model parameter to be significantly changed, while a more generous range for the bar size for a fixed prior range on C₀ returns an exponential disk with a larger scale length and a larger ellipticity that overpredicts the number of stars particularly in the outer disk. These larger disks with a higher ellipticity, however, are inconsistent with previous studies (e.g., van der Marel 2001; Saha et al. 2010; Balbinot et al. 2015; Mackey et al. 2016).

To achieve the computational efficiency and accuracy, we compute the log probability:

$$\begin{eqnarray}&&\mathrm{ln}P({\mu }_{m}\,| \,{\mu }_{\mathrm{obs}})\propto \mathrm{ln}P({\mu }_{\mathrm{obs}}\,| \,{\mu }_{m})+\mathrm{ln}P({\mu }_{m}),\end{eqnarray} \tag{ 10 }$$

where $\mathrm{ln}P({\mu }_{\mathrm{obs}}\,| \,{\mu }_{m})$ is a log-likelihood, and $\mathrm{ln}P({\mu }_{m})$ is the logarithm of the prior probability. For the stellar number count modeling, we use the Poisson likelihood:

$$\begin{eqnarray}\begin{array}{rcl}\mathrm{ln}P({\mu }_{\mathrm{obs}}\,| \,{\mu }_{m}) & = & \displaystyle \sum _{i}^{{nx}}\displaystyle \sum _{j}^{{ny}}[{\mu }_{\mathrm{obs},{ij}}\,\mathrm{ln}({\mu }_{m,{ij}})\\ & & -\,{\mu }_{m,{ij}}-\mathrm{ln}({\mu }_{\mathrm{obs},{ij}}!)].\end{array}\end{eqnarray} \tag{ 11 }$$

Instead of starting the walkers randomly sampled from flat priors, we run MCMC for 100–1000 steps several times to find a good initial guess for parameters. Restarting the walkers around the maximum log-probability position of the last step in the previous run is an efficient way to find reasonable initial guesses of the parameters. This technique also helps prevent the walkers from getting stuck at local likelihood maxima.

Once we set a reasonable initial point in the parameter space, we start 100 independent walkers in a small 11-dimensional sphere around the initial guess. At earlier steps in the chain, the walkers do not represent the true pPDFs. Once we run MCMC long enough (∼10 autocorrelation times; Foreman-Mackey et al. 2013), the ensemble of the walkers approximates the true pPDFs and thus we can estimate the parameters after discarding the burn-in steps in the chain. We determine conservatively the length of the burn-in phase based on the log probability over steps in the chain to assure the minimal effect of initial values.

In the first scenario, the ring-like overdensity formed from the evolution of a one-armed spiral. A one-armed spiral, triggered by repeated tidal encounters with the SMC, can wind up around the LMC main body over time (∼1–2 Gyr, e.g., Gómez et al. 2013; Besla et al. 2016). The amplitude of this structure may have become more pronounced after the SMC's recent collision (∼150 Myr ago, Zivick et al. 2018). This picture is consistent with simulations of repeated interactions between the MCs (Besla et al. 2012). Theoretical simulations of this general scenario have illustrated that a tidally induced ring-like spiral overdensity can be 2–3× larger than the smooth disk (e.g., Purcell et al. 2011; Gómez et al. 2013), which agrees with the amplitude of the observed structure.

In this scenario, the lack of young stellar populations can be naturally understood in that the disturbances induced by the SMC at earlier times primarily affect the outer disk, which is dominated by older stars and likely has no sufficient dense gas to form stars. Even if star formation was accompanied with the formation of a spiral arm by earlier (more than 1 Gyr ago) encounters, the continuation of star formation to the present day might be difficult owing to the very low gas density in the outskirts of the LMC disk. The mean H i column density decreases rapidly beyond 2.5 kpc and drops below N(H i) = 10²⁰ cm⁻² beyond 5 kpc (Staveley-Smith et al. 2003), indicating that the outer LMC disk is surrounded by very low density H i gas. There are only six young stars identified in the outskirts of the LMC (Moni Bidin et al. 2017), clearly showing that stars barely form out of this low gas density. The absence of young MS stars also has been reported in other galaxies' outer disks where the gas density is lower than 10²⁰ cm⁻² (=1 M_⊙ pc⁻²) (e.g., Grossi et al. 2011; Radburn-Smith et al. 2012; Hunter et al. 2013). Furthermore, the star formation histories measured in the LMC disk exhibit outside-in quenching of star formation; star formation activity has propagated inward over the last ∼1 Gyr (Meschin et al. 2014). This suggests that the LMC gas disk has significantly shrunk in the past gigayear. Secular gas depletion causes contraction of the star-forming disk over time in most dwarf galaxies (e.g., Stinson et al. 2009), and thus can be expected in the LMC as well.

However, external processes such as ram-pressure-stripping and tidal interaction with the SMC might be also important gas-stripping mechanisms for the LMC outskirts. The LMC gas disk has experienced ram-pressure-stripping since the MCs first entered into the virial radius of the MW (∼1 Gyr ago; Besla et al. 2012). The sharp falloff of the gas density beyond 4 kpc in the northeast leading edge, which is the moving direction of the LMC, with an extended old stellar disk well beyond ∼5 kpc, is clear evidence of ram-pressure-stripping (Salem et al. 2015). Repeated LMC–SMC interactions may strip gas from the LMC as well as the SMC (e.g., Nidever et al. 2008; Richter et al. 2013; Pardy et al. 2018). Thus, it is likely that the combination of all these processes significantly decreased the gas density in the outskirts over the last 1–2 Gyr, and the resulting low gas density prevented density waves (if the ones that are associated with the overdensity exist) from triggering shock-induced star formation in the overdensity during the past 1 Gyr. Or the strength of a shock could be too weak to form molecular clouds by compressing diffuse gas entering into a spiral arm.

In the second scenario, the ring-like overdensity is an immediate tidal response to the recent collision with the SMC (e.g., See Figure 3 in Athanassoula 1996). In this scenario, the structure can be interpreted as "ringing" (e.g., Gómez et al. 2012a, 2012b). Gómez et al. (2013) showed that radial density waves can induce a significant overdensity when a satellite galaxy plunges through the disk of its host galaxy. Although their simulation was optimized for the interaction between the MW and the Sagittarius dwarf spheroidal galaxy (with roughly a 1:10 mass ratio), the same logic can be applied to the LMC and SMC interaction, which have a similar mass ratio. Indeed, pronounced disturbances are expected in the LMC disk after the recent SMC collision (e.g., Besla et al. 2012; Pardy et al. 2016). In this scenario, recent star formation may be expected in the ring-like overdensity. However, there is no evidence of either ongoing star formation traced by Hα emission (SHASSA; Gaustad et al. 2001) or recent star formation traced by UV (GALEX; Martin et al. 2005) and far-IR (AKARI; Murakami et al. 2007) in the ring-like structure. If the overdensity is a simple stellar response to the density waves, the age of the composing stars is not necessarily connected to the age of the structure. The fact that the overdensity only shows up in intermediate-age or older stars may simply indicate that there was no dense gas to form new stars or no pre-existing young stars at those radii at the time of the recent collision. Due to the gas removal processes discussed in the previous section, it is likely that there was actually no sufficient dense gas left to form stars in the outer disk by the time of the recent collision (∼150 Myr ago).

In both scenarios, the resulting structures are kinematically induced and thus expected to be short-lived (∼1–2 Gyr; Berentzen et al. 2003; Yozin & Bekki 2014; Pardy et al. 2016), although repetitive encounters can reform the structure. Furthermore, in both scenarios, the origin of the ring-like overdensity at 6° is a product of tidal interactions with the SMC, rather than the MW. To test the origin of this structure, additional observational and numerical studies should be conducted. Observationally, we can look for (1) radial propagation either in the stellar motions themselves or older star formation across the ring overdensity as in Choi et al. (2015); and (2) radial variations of the mean vertical stellar velocity, which is expected in vertical density waves (Gómez et al. 2016, 2017). If one detects radial propagation in the stellar motions or radial variations in the vertical stellar velocity, this would indicate that the overdensity is likely a product of the recent collision. Numerically, the existence of the ring-like overdensity and its properties (position, amplitude, and shape) can be tested in high-resolution simulations by investigating different values of the LMC–SMC mass ratio, impact parameter, and timing of the collision. Because the position and amplitude of the ring-like overdensity strongly depend on these parameters, and in particular its impact parameter with the LMC, our findings will strongly constrain the recent (<500 Myr) dynamical evolution of the MCs. Given the sensitivity of the interaction history of the MCs to the tidal field of the MW, by including their kinematic information (e.g., Gaia Collaboration et al. 2018; Niederhofer et al. 2018; Zivick et al. 2018) we can also constrain the total dark matter halo mass of the MW—a low-impact-parameter collision between the MCs is less likely as the mass of the MW increases (Zivick et al. 2018). Thus, searching through the parameter space to simultaneously reproduce the observed MCs' morphology, kinematics, and its large-scale gas structure will constrain both dynamical models for the MCs and the halo mass of the MW.

Finally, these advances will help improve our general understanding of galaxy dynamics and the morphological evolution of interacting dwarf galaxies, using the MCs as a prototype for an interacting/colliding pair of galaxies. Dwarf–dwarf interactions are ubiquitous, but the morphological consequence of mergers between two small galaxies has not been fully investigated. Because the MCs' dynamics are complicated by their proximity to the MW, comparison with dwarf pairs in the field will enable us to better disentangle the environmental influence of a massive host versus dwarf–dwarf tidal effects (Stierwalt et al. 2015; Pearson et al. 2016). LMC–SMC analogs are rare in the field (<1%; Besla et al. 2018), but nearby examples exist. Magellanic Irregulars, NGC 4027 and NGC 3664, each have an SMC-mass companion. Both NGC 4027 (Phookun et al. 1992) and NGC 3664 (Wilcots & Prescott 2004) have one prominent spiral, an off-centered bar, and a smaller companion at ∼25–30 kpc from them. Even a sub-dominant spiral feature was detected in NGC 4027 (Phookun et al. 1992). Despite all these similarities to the LMC, there are two main differences: their completely wrapped ring contains both young and old stars; and their gas disks extend well beyond their stellar disks. This might indicate that the ring-like structure can be an evolutionary feature in interacting Magellanic-type galaxies.