A panoramic view of the Local Group dwarf galaxy NGC 6822

Dwarf galaxies are the most common type of galaxy in the Universe, and the building blocks of larger systems. As such, they play a crucial role as probes of the astrophysics of galaxy formation and evolution, especially in the context of the prevailing $\Lambda$ plus Cold Dark Matter ($\Lambda$CDM) framework (see e.g., Bullock & Boylan-Kolchin, 2017). Dwarf galaxies in the Local Group are particularly important to this endeavour, as they are sufficiently close that they may be studied, both photometrically and spectroscopically, on a star-by-star basis.

Because the halo mass function in $\Lambda$CDM is scale free, all dark matter halos that are large enough to host galaxies should be surrounded by substructures (e.g., Moore et al., 1999; Diemand et al., 2007; Springel et al., 2008). This is true both for giant galaxies like the Milky Way, as well as much smaller systems (e.g., Diemand et al., 2008; Li & Helmi, 2008). The practical consequence is that we should expect to see dwarf galaxies hosting systems of even smaller satellites (e.g., Wheeler et al., 2015). Moreover, we might expect to see low-mass stellar halos surrounding dwarf galaxies, since the build-up of such structures is at least partially due to the accumulation of tidally disrupted satellites over cosmic time (e.g. Bullock & Johnston, 2005; Cooper et al., 2010).

Experimental verification of these predictions is difficult. Even around large galaxies such as the Milky Way and Andromeda, stellar halos are extremely faint (e.g., Cooper et al., 2010; Ibata et al., 2014), and any such structure around a dwarf galaxy might reasonably be expected to have a commensurately lower luminosity. Unsurprisingly, evidence for their existence is therefore scarce (e.g., Minniti & Zijlstra, 1996; Minniti et al., 1999; Martínez-Delgado et al., 2012; Amorisco et al., 2014; McMonigal et al., 2016). Similarly, while a few Magellanic-type dwarfs in the Local Volume are known to host companions (e.g., Carlin et al., 2016, 2021), the mass scale for "satellites of satellites" is concentrated in the regime of ultra-faint systems (Wheeler et al., 2015), which have so far proven largely impossible to detect outside the Milky Way (for example, all known dwarfs around M31 have $M_{V}\leq-6$; Martin et al., 2016).

The Large and Small Magellanic Clouds (LMC/SMC), as the most massive dwarf satellites of the Milky Way, and among the closest, are a natural location to try and test the above ideas. However, even in these systems the observations are challenging – both galaxies span large angles on the sky such that very deep and wide photometric surveys are required. Numerous studies have searched without success for a stellar halo surrounding the LMC, although a multitude of perturbations to its outer disk are evident (e.g., Saha et al., 2010; Balbinot et al., 2015; Mackey et al., 2016, 2018; Belokurov & Erkal, 2019; Gaia Collaboration et al., 2021; Cullinane et al., 2021). Nonetheless, the LMC may possess a kinematically hot component (Minniti et al., 2003); moreover, Belokurov & Koposov (2016) photometrically detected a number of streamlike features in the extreme outskirts of the Clouds (at radii $25-50$ kpc, well beyond the structures seen in the LMC outer disk). Subsequent spectroscopic observations confirmed that at least two of these are plausibly associated with the Magellanic system (Navarrete et al., 2019), suggesting that it may indeed be surrounded by an exceptionally low-luminosity stellar envelope. The difficulty in detecting, characterising, and interpreting such a structure is compounded by the fact that the LMC and SMC are interacting both with each other and with the Milky Way; at a minimum, contamination by the Galactic stellar halo is a significant problem (Navarrete et al., 2019).

Despite the difficulty in determining the presence of a halo-like component around the Magellanic Clouds, the existence of Magellanic satellites is well established. Many of the ultra-faint dwarf galaxies uncovered in recent years by deep wide-field surveys of the southern hemisphere (e.g., Drlica-Wagner et al., 2015; Koposov et al., 2015; Bechtol et al., 2015) have been convincingly demonstrated to be members, or ex-members, of the Magellanic system (e.g., Jethwa et al., 2016; Dooley et al., 2017b; Sales et al., 2017; Kallivayalil et al., 2018; Erkal et al., 2020). However, again the Magellanic-Milky Way interaction complicates interpretation – association of satellites with the Clouds, or otherwise, must be done probabilistically, and depends on various assumptions such as the mass of the LMC (e.g., Erkal et al., 2020).

One way of mitigating the difficulties introduced by the fact that the LMC and SMC are interacting with the Milky Way is to move to more isolated targets. NGC 6822, sometimes known as Barnard’s Galaxy (Barnard, 1884), is the most massive of the handful of systems that sit just outside the Milky Way’s virial radius. It is a gas-rich dwarf irregular galaxy with a stellar mass $M_{*}\approx 10^{8}M_{\odot}$ (McConnachie, 2012), perhaps a quarter that of the SMC, plus in excess of $\sim 10^{8}M_{\odot}$ of Hi gas (e.g., de Blok & Walter, 2000). As such, it is an excellent location to search for low-mass satellites (e.g., Wheeler et al., 2015; Dooley et al., 2017a), and/or a stellar halo component.

Many previous studies have revealed NGC 6822 to be a fascinating object in its own right. It possesses both an extended elliptical spheroid composed of old stars, and an irregular disk-like Hi distribution hosting active star formation (e.g., de Blok & Walter, 2000, 2003, 2006; Komiyama et al., 2003; Demers et al., 2006; Battinelli et al., 2006; Fusco et al., 2014; Higgs et al., 2021). The main axis of the Hi disk is angled at approximately $60\degr$ to the major axis of the spheroid, leading to speculation that NGC 6822 might have experienced a recent merger event or tidal interaction, or even be a polar ring galaxy (e.g., de Blok & Walter, 2000, 2006; Demers et al., 2006; Battinelli et al., 2006).

However, Hubble Space Telescope imaging at various locations in the system by Cannon et al. (2012), including the putative companion galaxy, revealed no obvious variation in stellar populations, leading these authors to suggest that the Hi distribution most likely reflects a warped disk inclined to the line of sight. This idea is supported by the fact that the spheroid is observed to exhibit prolate rotation (i.e., about its major axis, Thompson et al., 2016; Belland et al., 2020), which is around approximately the same axis as the rotation of the Hi disk as measured by, e.g., Weldrake et al. (2003). NGC 6822 also hosts a globular cluster system numbering $8$ objects (e.g., Hubble, 1925; Hwang et al., 2011; Huxor et al., 2013). This system is unusual in that the clusters are largely distributed along the major axis of the spheroid (see Figure 1); it may also exhibit weak net rotation, albeit with poorly constrained orientation (Veljanoski et al., 2015).

Several recent studies have attempted to measure a proper motion for NGC 6822 using astrometry from the Gaia satellite (e.g., McConnachie et al., 2021; Battaglia et al., 2021). The implied orbital parameter space indicates that NGC 6822 has likely been an isolated system for the majority of its life; however, there is a non-negligible chance that it has passed within the virial radius of th Milky Way in the past few Gyr. In this case, NGC 6822 may constitute a "backsplash" galaxy – such systems are expected to be relatively common around giant galaxies like the Milky Way (e.g., Gill et al., 2005; Teyssier et al., 2012; Buck et al., 2019).

In this paper, we present deep panoramic imaging of NGC 6822, extending to a maximum radius of $\sim 2\degr$ from the centre of the dwarf. This is a much wider area than covered by previous surveys. Our main aim is to investigate the extreme outskirts of this system for the first time, searching for (i) stellar streams or overdensities, or a halo-like component that may indicate recent merger activity; (ii) tidal disturbances that may indicate a previous interaction with the Milky Way; and (iii) low-luminosity satellites. Throughout this work we assume that the centre of NGC 6822 is at $\alpha=19^{\rm h}\,44^{\rm m}\,56.6^{\rm s},\delta=-14\degr\,47\arcmin\,21\arcsec$ (McConnachie, 2012), and that it sits at a distance of $472$ kpc (derived as a weighted average of the Cepheid, RR Lyrae, and tip of the red giant branch (TRGB) distances compiled in Górski et al., 2011). At this distance, one arcsecond subtends $\approx 2.29$ pc.

A by-eye analysis of our spatial density maps (Figures 5 and 6) reveals no strong density peaks that might indicate the presence of one or more satellites of NGC 6822. Given that we can detect several of the globular clusters in the system via resolved star counts, at least with our Megacam data, we expect that if any dwarfs were present they should be visible in these maps down to some faint limit. In this section we aim to determine this faint limit in order to place quantitative constraints on the absence of satellite galaxies in the NGC 6822 system.

To do this, we insert artificial dwarfs into our data, spanning a grid of size and luminosity to identify the region of parameter space where a satellite is no longer visible in our maps. We consider luminosities in the range between $M_{V}=-10$ and $-4$. For each artificial dwarf we randomly generate stars according to a model from the Dartmouth Stellar Evolution Database (Dotter et al., 2008). We assume an ancient metal-poor population with age $12$ Gyr and metallicity [Fe/H]$=-2.1$, which is appropriate for dwarfs in this luminosity range (see the luminosity-metallicity relation described by e.g., Kirby et al., 2013; Simon, 2019). We further assume [$\alpha$/Fe]$=+0.2$, which is appropriate for faint dwarfs in the Milky Way.

Given these specifications, the Dartmouth models provide the number of stars in different stellar luminosity bins according to a Salpeter (1955) initial mass function (IMF). This allows us to determine the number of observable stars ($N^{*}$) in a given artificial dwarf, by applying our pan-mosaic faint limits for either the Megacam or DECam data. This process is complicated by the fact that the Dartmouth luminosity functions do not include evolution past the helium flash. We therefore need to correct our calculations for missing sources on the horizontal branch and asymptotic giant branch, which typically number $\sim 25\%$ of the observed RGB population in an old stellar system.

We distribute the observable stars spatially by assuming that the radial density profile for each artificial dwarf follows the simple exponential form proposed by Martin et al. (2016):

where $e$ is the ellipticity of the dwarf, $r$ is the distance along its major axis, and $r_{h}$ is its half-light radius. The compilation by Simon (2019, their Figure 2) shows that for dwarf satellites of the Milky Way, the half-light radius ranges from $r_{h}\approx 500$ pc for objects at the upper end of our luminosity range ($M_{V}\sim-10$) down to $r_{h}\approx 50$ pc for objects at the lower end of the range ($M_{V}\sim-4$). We therefore consider three characteristic sizes of $r_{h}=50$, $100$, and $500$ pc. For a given artificial dwarf we further select an ellipticity randomly from a uniform distribution spanning the range $0\leq e\leq 0.6$³³3Simon (2019) shows that the mean ellipticity for classical dwarf satellites of the Milky Way is $\approx 0.35$ and for the ”ultra-faint” satellites is $\approx 0.5$., and a random position angle ($\theta$) for the major axis.

Given these parameters, and the above equation, for each of the $N^{*}$ observable stars in an artificial dwarf we randomly generate a radius $r$ along the major axis, and a random position angle relative to the major axis. We place the dwarf at some location relative to the centre of NGC 6822, which allows us to calculate $(x,y)$ positions for all stars in the galaxy-centric coordinate system used for our density maps. It is then straightforward to add these stars to our catalogue and remake the map. In total we selected four different sites for each of the Megacam and DECam mosaics, each of them outside several scale lengths of the NGC 6822 spheroid (i.e., outside $\sim 30-40\arcmin$).

An example of the appearance of an artificial dwarf in the DECam map is shown in Figure 10. With a luminosity $M_{V}=-6$ and half-light radius $r_{h}=100$ pc, this object is prominent and would have been evident by eye almost anywhere in the DECam mosaic. We quantified the significance the density peak due to each dwarf in units of the standard deviation ($\sigma$) of map pixels surrounding the dwarf’s location. For the example in Figure 10 the significance is $\sim 10\sigma$; in the literature, detection significance thresholds of $\sim 5-10\sigma$ are commonly utilised (e.g., Koposov et al., 2015; Drlica-Wagner et al., 2015; Kim & Jerjen, 2015; Kim et al., 2015).

Figure 11 shows the detection significance for artificial dwarfs of different size and luminosity across our DECam and Megacam mosaics. In general, these objects have lower detection significance in the DECam map compared to the Megacam map, because the DECam photometry is shallower (i.e., for a given dwarf luminosity there are fewer observable stars). Moreover, there is a clear size dependence – at fixed luminosity the more compact systems produce a stronger central peak resulting in a higher detection significance. Figure 11 demonstrates that the most diffuse dwarfs, with $r_{h}=500$ pc, have become marginally detectable by $M_{V}\sim-6.5$, while more compact systems reach a $5\sigma$ threshold between $-4.8\leq M_{V}\leq-5.5$. There is a small amount of variation from location to location, but in general the results have relatively small scatter.

In summary, the faintest detectable satellite dwarfs in our data would have $M_{V}\sim-5$ and half-light radii smaller than $\sim 100$ pc. More diffuse objects would need to be brighter, $M_{V}\sim-6.5$. Another way of thinking about this is in terms of the limiting central surface brightness – for a dwarf with $M_{V}=-5$ and $r_{h}=100$ pc at the distance of NGC 6822 the central surface brightness is $V\approx 28.6$ mag arcsec^-2, while for a system with $M_{V}=-6.5$ and $r_{h}=500$ pc it is fainter than $V\sim 30$ mag arcsec^-2. These values are in reasonable agreement with our previous estimate for the faint surface brightness limit of our survey (see Section 3.2).

In physical terms our detection thresholds are relatively high – many fainter satellites are known around the Milky Way, that we would not be sensitive to. This includes all the proposed ultra-faint companions of the Magellanic Clouds (e.g., Erkal et al., 2020). The predictions of Wheeler et al. (2015) suggest a $\sim 35\%$ chance of a galaxy like NGC 6822 possessing a dwarf satellite with stellar mass above $3000M_{\odot}$ within its virial radius (which could be as large as $\sim 50$ kpc). More encouragingly, the abundance-matching methods from Dooley et al. (2017a) suggest that NGC 6822 could harbour up to $\sim 4$ satellites with stellar masses greater than $10^{4}M_{\odot}$. Assuming a mass-to-light ratio of $\approx 1$ for an ancient metal-poor population, these two limiting stellar masses correspond to luminosities of $M_{V}\approx-3.9$ and $\approx-5.2$, respectively. The first of these is at least a magnitude fainter than our most optimistic detection limit, while the second sits near our faint limit. Importantly, our DECam data also only extend to less than half the expected virial radius of NGC 6822. Together, these results suggest that it would be beneficial for future deeper and even wider-field surveys – such as the Legacy Survey of Space and Time (LSST) to be conducted by the Vera Rubin Observatory – to revisit the question of satellites around this isolated system.

One way of explaining the observed flattening and twisting of the outer spheroid in NGC 6822 might be if this galaxy had experienced a weak gravitational disturbance at some point in the past. The availability of Gaia Early Data Release 3 (EDR3, Gaia Collaboration, 2021) opens the possibility of measuring a proper motion for NGC 6822. This, combined with its position, and estimates of its distance and line-of-sight velocity, in turn allows us to calculate some simple orbits in order to test the extent to which NGC 6822 has been a truly isolated system.

Previously, McConnachie et al. (2021) used Gaia Data Release 2 (DR2, Gaia Collaboration et al., 2018) to address this problem, finding a systemic proper motion of $(\mu_{\alpha},\,\mu_{\delta})=(-0.02\pm 0.02,\,0.00\pm 0.02)$ mas/yr. Here the reported uncertainties are random only, and in fact the measurement is dominated by systematics at the level of $\approx 0.035$ mas/yr. Nonetheless, McConnachie et al. (2021) concluded that NGC 6822 has likely fallen in towards the Milky Way from the edge of the Local Group, but has not had an interaction with the Milky Way (i.e., has not passed within its virial radius of $\sim 300$ kpc) at $\approx 90\%$ confidence. Similarly, their orbit sampling suggested that NGC 6822 has not had an interaction with M31 at $\approx 99.99\%$ confidence.

On angular scales of a few degrees, Gaia EDR3 has systematic proper motion uncertainties at least a factor of two smaller than Gaia DR2 (e.g., Lindegren et al., 2021), which makes it worthwhile revisiting this analysis. We first downloaded from the Gaia science archive a list of all sources in EDR3 sitting within $0.3$ degrees of the centre of NGC 6822. We then used a series of cuts to isolate young members of NGC 6822, and keep only stars with high quality astrometry. We selected young stars as our tracers because we showed above (Section 3.2) that members of this population are affected by contamination at a much lower level than RGB and red clump stars. In addition, young supergiants reach brighter magnitudes than stars on the upper RGB, which increases the chance of selecting a few members with excellent astrometry.

To isolate the young population in NGC 6822, we selected stars with colour $-0.4<G_{\rm BP}-G_{\rm RP}<0.7$ and magnitude $G<20$. We further demanded that the parallax $|\varpi|<2\sigma_{\varpi}$, where $\sigma_{\varpi}$ is the error in the parallax. We limited $\sigma_{\varpi}<0.5$ mas to ensure we only kept stars with fairly reliable measurements.

Because the centre of NGC 6822 is very crowded, many stars possess poor astrometry. To try and exclude such sources we made a few additional cuts. First, as recommended by Gaia Collaboration (2021), we limited the parameter RUWE (Renormalised Unit Weight Error) to be RUWE $<1.4$, as sources with RUWE greater than this are typically unresolved binaries or blends. Second, we limited astrometric_excess_noise $<1.5$, which, again, is useful for eliminating stars that are blends or binaries (e.g., Iorio & Belokurov, 2019). Finally, following Vasiliev (2019), we selected only stars with phot_bp_rp_excess_factor $<1.3+0.06(G_{\rm BP}-G_{\rm RP})^{2}$, which helps further reduce the number of blended sources.

Figure 12 shows how this process improves the selection of NGC 6822 members. After our quality cuts, the remaining sources clearly trace the expected on-sky distribution of young stars in NGC 6822, and produce a very concentrated distribution on the proper motion plane. Altogether our final sample comprises $266$ stars.

To measure the proper motion of NGC 6822, we fit a 2D Gaussian to the stellar distribution on the proper motion plane. Based on our earlier examination of the young stellar populations in NGC 6822 we assume that the level of contamination by non-members is low, and elect not to fit a background contribution together with the Gaussian. However, to ensure that proper motion outliers do not affect the fit, we calculate the median and standard deviation for each component across the sample, and retain only objects within $\pm 3\sigma$ of the median for the final fit.

We use a maximum likelihood method to determine the most likely parameters for the 2D Gaussian given the observed stellar positions on the proper motion plane. The Gaussian has a peak at $(\mu_{\alpha,0},\,\mu_{\delta,0})$, and a width defined by $\sigma_{\mu_{\alpha}}$ and $\sigma_{\mu_{\delta}}$ plus a covariance $\rho$. The latter three parameters help account for the internal kinematics of young stars in NGC 6822, and could be non-zero in the case that the EDR3 astrometry is sufficiently good to resolve such motions. For reference, Kirby et al. (2014) find that the line-of-sight velocity dispersion for (old) RGB stars in the spheroid is $23.2$ km/s, which would equate to a proper motion of $\sim 0.01$ mas/yr. However, the dispersion of the young populations is unknown, and plausibly substantially smaller given the likelihood of some degree of rotational support (which is seen for the Hi).

We sample the posterior probability distributions for the parameters in our model using the Markov Chain Monte Carlo sampler emcee (Foreman-Mackey et al., 2013). We assume uninformative priors for all parameters. When performing the fit, we took into account the individual proper motion uncertainties per star from Gaia, as well as the covariance between the Gaia measurements of the two proper motion components for each star.

Figure 13 shows the one- and two-dimensional marginalized posterior probability distribution functions for the five parameters in the model. From this plot it is evident that the proper motion for NGC 6822 is clearly measured and non-zero: we find $\mu_{\alpha,0}=-0.061\pm 0.014$ mas/yr and $\mu_{\delta,0}=-0.083\pm 0.012$ mas/yr.

We also find that the dispersion parameters are essentially unresolved, although both exhibit small tails to non-zero values that suggest the data are on the cusp of being able to provide information about the internal motions of stars in NGC 6822. It would be worthwhile revisiting this possibility in future Gaia data releases.

To check the proper motion zero-point in this region of the sky, we follow van der Marel et al. (2019) and download astrometry for quasars using the agn_cross_id table available in the Gaia science archive. Because these sources are relatively sparsely distributed, we needed to expand our search radius to $\sim 6\degr$ from the centre of NGC 6822 in order to locate a sufficient number. We then applied the same cuts described above (excluding the colour-magnitude selection), and again fit a 2D Gaussian model on the proper motion plane. This has a peak at $\mu_{\alpha,{\rm ZP}}=0.008\pm 0.012$ mas/yr and $\mu_{\delta,{\rm ZP}}=0.006\pm 0.010$ mas/yr. We subtract this from our best-fit value for NGC 6822 to obtain a final systemic motion of $\mu_{\alpha,0}=-0.069\pm 0.019$ mas/yr and $\mu_{\delta,0}=-0.089\pm 0.016$ mas/yr (where the uncertainties have been combined in quadrature).

Our proper motion measurement is mildly inconsistent with the Gaia DR2 value obtained by McConnachie et al. (2021), even considering the substantial systematics that may affect their measurement. This is especially notable in the $\mu_{\delta}$ component. However, while we were writing this paper, Battaglia et al. (2021) released a set of uniform Gaia EDR3 proper motion measurements for galaxies across the Local Group. They find $\mu_{\alpha,0}=-0.06\pm 0.01$ mas/yr and $\mu_{\delta,0}=-0.07\pm 0.01$ mas/yr, derived using both young and old stellar populations and including a background contribution in their fit. They also measured a quasar zero-point of $\mu_{\alpha,{\rm ZP}}=-0.008\pm 0.020$ mas/yr and $\mu_{\delta,{\rm ZP}}=-0.016\pm 0.020$ mas/yr. Both these results are much more compatible with the results presented here.

Having determined a proper motion for NGC 6822, we aim to briefly investigate its orbit relative to the Milky Way. To do this we use the galpy galactic dynamics package (Bovy, 2015) to integrate the motion of NGC 6822 backwards in time for $10$ Gyr. We assume the usual position for the centre of the galaxy at $\alpha=19^{\rm h}\,44^{\rm m}\,56.6^{\rm s},\delta=-14\degr\,47\arcmin\,21\arcsec$ (McConnachie, 2012), our previously-adopted distance of $472$ kpc, and the heliocentric line-of-sight velocity of $v_{\rm los}=-54.5\pm 1.7$ km/s measured by Kirby et al. (2014). The orbit is integrated in the Milky Way potential from McMillan (2017), which has a total mass of $1.3\times 10^{12}\,M_{\odot}$.

The top panel of Figure 14 shows the distance of NGC 6822 from the centre of the Milky Way as a function of lookback time. NGC 6822 has fallen inwards from a distance of nearly $600$ kpc, and appears to have had a weak interaction with the Milky Way, passing just inside its virial radius (assumed to be $\sim 300$ kpc) approximately $4$ Gyr ago. The orbit is bound – integration further back in time reveals an apocentre of $\approx 600$ kpc roughly $12$ Gyr ago.

These conclusions are sensitive to measurement uncertainties as well as several of our assumptions. To check the effect of measurement uncertainties we randomly generated a set of $10^{5}$ new distances, proper motions, and line-of-sight velocities and then re-integrated the orbit for each combination. For the proper motions and line-of-sight velocity, we drew from normal distributions centred on the values outlined above and with widths defined by the reported uncertainty intervals. For the distance, we assumed a range of $\pm 0.08$ mag on the distance modulus, which is commensurate with typical uncertainties reported in the literature (cf. Górski et al., 2011; McConnachie, 2012; Higgs et al., 2021).

The upper panel of Figure 14 shows $100$ example orbits from the randomly generated set. A wide variety of different trajectories are permissible given the current measurement uncertainties, ranging from orbits where NGC 6822 has fallen in from a very large distance and is essentially only now at its closest approach to the Milky Way, to orbits where NGC 6822 plunges deep within the virial radius of the Milky Way and has had a relatively strong encounter. The latter type of orbit also tends to be bound to the Milky Way, with apocentres around $\sim 500$ kpc; however, we note that long-term integration of this type of orbit may not be reliable as our simple model does not account for dynamical friction, or changes in the mass of the Milky Way with time. Across the set of $10^{5}$ realisations, $\approx 57\%$ pass within $300$ kpc of the Milky Way; the median time of pericentre for these encounters is $3.1$ Gyr ago. Around $27\%$ of orbits are bound to the Milky Way.

Of the four sampled parameters, uncertainties on the distance are most important. Indeed, our assumed error of $\pm 0.08$ mag may be an under-estimate – this range includes the shorter distance of $459\pm 17$ kpc listed by McConnachie (2012) (originally from Gieren et al., 2006) as well as the larger distance of $490$ kpc from Mateo (1998) and the majority of measurements compiled in Górski et al. (2011); however, recently Higgs et al. (2021) obtained a much higher estimate of $570\pm 12$ kpc using the RGB tip. The lower panel of Figure 14 shows the orbit derived using this distance together with the velocities outlined above (red line). In this case, NGC 6822 does not approach the Milky Way any closer than $\approx 500$ kpc.

The assumed gravitational potential of the Milky Way is also important. Because NGC 6822 is quite distant, it is not generally the shape of the potential that matters, but rather the overall assumed mass. As noted above, the McMillan (2017) potential has a Milky Way mass of $1.3\times 10^{12}\,M_{\odot}$; the lower panel of Figure 14 shows orbits integrated in the same potential but with smaller ($0.8\times 10^{12}\,M_{\odot}$; green line) and larger ($2.0\times 10^{12}\,M_{\odot}$; blue line) halo masses. Perhaps unsurprisingly, for a higher-mass Milky Way the orbit reaches more deeply within the virial radius, while for a lower-mass Milky Way the orbit does not cross the virial radius.

It is worth explicitly pointing out that our simple models do not include the effect of other galaxies such as M31 or the LMC. McConnachie (2012) showed that in general M31 remains very distant from NGC 6822 regardless of the assumed initial conditions, and we find a similar outcome here (see also Battaglia et al., 2021). On the sky, NGC 6822 does not pass within an angular distance of $\sim 50\degr$ of the current position of M31; because M31 is on a rather radial orbit (van der Marel et al., 2012, 2019) this position does not change much with time. At the current distance of M31 an angular separation of $>50\degr$ corresponds to a projected physical distance of at least $\sim 680$ kpc. Because M31 is currently moving towards the Milky Way (and was hence further away in the past) this is a lower limit. The LMC is more difficult to account for; indeed Battaglia et al. (2021) show that it might affect the orbit of NGC 6822 during the past $\sim 3-4$ Gyr. More detailed investigation is clearly warranted.

Our finding that NGC 6822 may have passed within the virial radius of the Milky Way is consistent with recent results from both McConnachie et al. (2021) and Battaglia et al. (2021). The former find that NGC 6822 can pass within the Milky Way’s virial radius, but observe such an encounter in fewer than $10\%$ of their sampled orbits. However, as noted above their measured proper motions are somewhat different than ours. On the other hand Battaglia et al. (2021), whose measurements are entirely consistent with ours, find that NGC 6822 could certainly have passed within the virial radius of the Milky Way; indeed for their "heavy Milky Way" model the pericentre is as close as $100$ kpc.

In summary, we measure a clear non-zero proper motion for NGC 6822 using Gaia EDR3. Integrating the orbit of this galaxy, we find that it has most likely fallen in towards the Milky Way, from a direction largely away from that of M31, and a distance of at least $500$ kpc. There is a good chance that NGC 6822 passed within the virial radius of the Milky Way $\sim 3-4$ Gyr ago; if so, it is likely the best local example of a "backsplash" galaxy. It is also possible that NGC 6822 is on a very long-period bound orbit about the Milky Way. A previous weak gravitational interaction with our Galaxy might help explain the twisting and flattening of the outer spheroid that we have observed in our imaging data. Moreover, the star-formation history derived by Weisz et al. (2014) (see also Cannon et al., 2012) suggests that NGC 6822 formed only about $\sim 30\%$ of its stars earlier than $6$ Gyr ago, and the star-formation rate shows a substantial acceleration since $\sim 3-5$ Gyr ago (somewhat location-dependent). This may also be consistent with a scenario where NGC 6822 interacted with the Milky Way around this time.

We have used Blanco/DECam and Magellan/Megacam imaging to conduct a panoramic survey of the isolated Local Group dwarf galaxy NGC 6822. Our photometry reaches $\sim 2-3$ magnitudes deeper than most previous studies and we cover the widest area around the dwarf compared to any prior work.

We used young main sequence stars, and old stars from the red giant branch and red clump, to trace the distribution of stellar populations in NGC 6822 with the following main results:

1. 1.

   NGC 6822 has substantial young stellar populations, which closely trace the Hi gas in the dwarf. This is consistent with the findings from many previous studies (e.g., Komiyama et al., 2003; de Blok & Walter, 2006). The Hi is known to be arranged in a rotating disk-like structure (e.g., de Blok & Walter, 2000). We do not observe any substantial young populations away from the gas.

2. 2.

   NGC 6822 possesses a large extended spheroidal component made up of older stellar populations (age $\geq 2$ Gyr). The major axis of the spheroid is angled by $\approx 60\degr$ to the direction of the Hi gas. Again, this is known from many previous studies (e.g., Komiyama et al., 2003; de Blok & Walter, 2006; Battinelli et al., 2006; Higgs et al., 2021). Our data clearly show that there is no correlation between dense clumps of young stars, and the density of older populations.

3. 3.

   We were able to trace the extent of the spheroid to a larger radius than any previous work, $\approx 60-80\arcmin$ ($8.2-11.0$ kpc). This is commensurate with the radial extent of the globular cluster system in NGC 6822. The spheroid appears smooth, and its density profile outside $\sim 10\arcmin$ is well fit by a single exponential function with scale radius $9.6\pm 0.3\arcmin$ (random) $\pm 0.5\arcmin$ (systematic) ($1.3\pm 0.1\pm 0.1$ kpc). We do not observe any break in the profile that would indicate the presence of a more extended stellar halo-like component surrounding NGC 6822.

4. 4.

   Both the ellipticity and orientation of the spheroid change with radius. The system becomes more flattened at larger radii, and the position angle of the major axis twists by up to $40$ degrees.

5. 5.

   We do not observe any stellar streams or over-densities to the edge of our DECam mosaic (radius $\approx 2\degr\sim 16.5$ kpc) and to a faint limit of $V\sim 30$ mag arcsec^-2. There are some low-level fluctuations visible in the density maps, but these cannot be robustly distinguished from background noise.

The structure of NGC 6822 is a long-standing puzzle, in particular the very different orientation between the spheroid and the Hi. Studies such as those of de Blok & Walter (2000, 2003) and Komiyama et al. (2003) suggested that the presence of the extended Hi structure and associated young populations could be due to a recent merger event, with the large cloud of gas to the north-west potentially indicating the location of the interacting system. Demers et al. (2006) and Battinelli et al. (2006) even suggested NGC 6822 could be a polar ring galaxy. However, the Hubble Space Telescope study of Cannon et al. (2012) showed that the stellar populations at the location of the north-west Hi cloud were indistinguishable from the stellar populations at other locations in the dwarf. They therefore suggested that the Hi structure in NGC 6822 is simply a warped disk inclined to the line-of-sight. The observed kinematic connection between the rotation of the spheroid and the rotation of the gas is consistent with this interpretation (e.g., Thompson et al., 2016; Belland et al., 2020).

Our observation that there is no over-density of old stellar populations at the location of the north-west gas cloud supports the conclusion of Cannon et al. (2012) that the north-west gas cloud is not an interacting companion galaxy. Moreover, we find that the spheroid appears mostly undisturbed (smooth) and observe no obvious stellar streams or over-densities of either young or old stars near to NGC 6822. This all indicates that NGC 6822 has probably not experienced any significant merger events in the past few Gyr. This may be consistent with recent hydrodynamical models which indicate that lower-mass dwarfs may largely form in place without requiring merger activity to shape their properties (e.g., Wheeler et al., 2019); however, we cannot exclude the possibility that NGC 6822 underwent one or more mergers early in its history. The origin of the substantial mis-alignment between the gas disk and spheroid major axes, reflected in the apparent prolate rotation of the spheroid (Belland et al., 2020), remains a mystery. Detailed modelling is required to assess whether such a configuration could arise from an early merger event; moreover, future astrometry from the Gaia satellite might resolve internal motions to help address this question. Data from Gaia EDR3 are on the edge of being able to achieve this.

We did not observe any compact over-densities that would indicate the presence of dwarf satellites of NGC 6822, out to the edge of our DECam mosaic at $\sim 16.5$ kpc. By inserting artificial dwarf galaxies into our smoothed density maps, we show that we could have detected any faint satellite of NGC 6822 down to a luminosity of $M_{V}\approx-6.5$ in the case of diffuse dwarfs ($r_{h}\approx 500$ pc) or $M_{V}\approx-5$ in the case of more compact satellites ($r_{h}\approx 50-100$ pc). This is consistent with the fact we were able to resolve a few of the better-populated globular clusters, at least with Megacam. If NGC 6822 has any low-luminosity satellites, they must be fainter than our detection limit or sit outside the surveyed area. Our data reach only to approximately half the expected virial radius of the system, so it would be worthwhile revisiting this question with future wide-field surveys such as LSST.

Finally, we compute the proper motion of NGC 6822 using astrometry from Gaia EDR3 and show that this galaxy has indeed likely spent most of its life in isolation, falling inwards from a large distance ($>500$ kpc) at the edge of the Local Group. This is in agreement with McConnachie et al. (2021). Our orbit calculations also strongly suggest that NGC 6822 passed within the virial radius of Milky Way $\sim 3-4$ Gyr ago, in agreement with recent results from Battaglia et al. (2021). A weak gravitational encounter of this nature might help explain the flattening and twisting we observed in the outer spheroid, as well as the vigorous increase in star formation seen in the last $\sim 3-5$ Gyr (e.g., Weisz et al., 2014).

We are grateful to Erwin de Blok for supplying his zero-moment Hi map for NGC 6822. DM acknowledges support from an Australian Research Council Future Fellowship (FT160100206).

This paper includes data gathered with the 6.5m Magellan Telescopes located at Las Campanas Observatory, Chile. Australian access to the Magellan Telescopes was supported through the National Collaborative Research Infrastructure Strategy of the Australian Federal Government. This paper uses data products produced by the OIR Telescope Data Center, supported by the Smithsonian Astrophysical Observatory.

This project used data obtained with the Dark Energy Camera (DECam), which was constructed by the Dark Energy Survey (DES) collaboration. Funding for the DES Projects has been provided by the US Department of Energy, the US National Science Foundation, the Ministry of Science and Education of Spain, the Science and Technology Facilities Council of the United Kingdom, the Higher Education Funding Council for England, the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, the Kavli Institute for Cosmological Physics at the University of Chicago, Center for Cosmology and Astro-Particle Physics at the Ohio State University, the Mitchell Institute for Fundamental Physics and Astronomy at Texas A&M University, Financiadora de Estudos e Projetos, Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro, Conselho Nacional de Desenvolvimento Científico e Tecnológico and the Ministério da Ciência, Tecnologia e Inovação, the Deutsche Forschungsgemeinschaft and the Collaborating Institutions in the Dark Energy Survey.

The Collaborating Institutions are Argonne National Laboratory, the University of California at Santa Cruz, the University of Cambridge, Centro de Investigaciones Enérgeticas, Medioambientales y Tecnológicas–Madrid, the University of Chicago, University College London, the DES-Brazil Consortium, the University of Edinburgh, the Eidgenössische Technische Hochschule (ETH) Zürich, Fermi National Accelerator Laboratory, the University of Illinois at Urbana-Champaign, the Institut de Ciències de l’Espai (IEEC/CSIC), the Institut de Física d’Altes Energies, Lawrence Berkeley National Laboratory, the Ludwig-Maximilians Universität München and the associated Excellence Cluster Universe, the University of Michigan, NSF’s NOIRLab, the University of Nottingham, the Ohio State University, the OzDES Membership Consortium, the University of Pennsylvania, the University of Portsmouth, SLAC National Accelerator Laboratory, Stanford University, the University of Sussex, and Texas A&M University.

Based on observations at Cerro Tololo Inter-American Observatory, NSF’s NOIRLab (NOIRLab Prop. ID 2013A-0617 and 2013B-0611; PI: D. Mackey), which is managed by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation.

The Pan-STARRS1 Surveys (PS1) and the PS1 public science archive have been made possible through contributions by the Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its participating institutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, the Queen’s University Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, the National Aeronautics and Space Administration under Grant No. NNX08AR22G issued through the Planetary Science Division of the NASA Science Mission Directorate, the National Science Foundation Grant No. AST-1238877, the University of Maryland, Eotvos Lorand University (ELTE), the Los Alamos National Laboratory, and the Gordon and Betty Moore Foundation.

The data underlying this article will be shared on reasonable request to the corresponding author.