Eridanus IV: an Ultra-faint Dwarf Galaxy Candidate Discovered in the DECam Local Volume Exploration Survey

DELVE is an ongoing observational campaign using the Dark Energy Camera (DECam; Flaugher et al. 2015) on the 4 m Blanco telescope to study the satellite systems of the Milky Way, Magellanic Clouds, and four Magellanic analogs within ∼ 2 Mpc (Drlica-Wagner et al. 2021). DELVE seeks to complete observational coverage of the high-Galactic-latitude (∣b∣ > 10°) southern sky in the g, r, i, z bands by combining 126 nights of observing with all existing public DECam community data with exposure times > 30 s. More detailed information about the observational strategy and scientific goals of DELVE can be found in Drlica-Wagner et al. (2021).

We used an updated internal data release from DELVE constructed from ∼ 20,000 g- and i-band exposures covering ∼6400 deg² in the southern Galactic cap outside the DES footprint. All exposures were processed with the DESDM pipeline (Morganson et al. 2018), which utilizes SourceExtractor and PSFEx (Bertin & Arnouts 1996; Bertin 2011) for automatic source detection and photometric measurement. Astrometry was calibrated against Gaia DR2, which provides ∼ 30 mas relative astrometric calibration. The DELVE photometry is calibrated for each DECam CCD by matching stars to the ATLAS Refcat2 catalog (Tonry et al. 2018). A multi-band catalog of unique astronomical objects was generated by matching the SourceExtractor catalogs from each CCD image as described in Drlica-Wagner et al. (2021). For each source, we calculated interstellar extinction due to foreground dust by interpolating the E(B − V) maps of Schlegel et al. (1998) with the rescaled normalization from Schlafly & Finkbeiner (2011) and the reddening coefficients from Section 4.2 of DES Collaboration (2018). Hereafter, all quoted magnitudes are corrected for extinction.

We separate stars and galaxies using the Source Extractor SPREAD_MODEL parameter. When performing our search, we selected a relatively pure sample of stars, ∣SPREAD_MODEL_G∣ < 0.003 + ∣SPREADERR_MODEL_G∣, and restricted our stellar sample to g < 23.0 mag, which corresponds to the median g-band magnitude limit of the DELVE data at S/N = 10. To characterize Eri IV, we selected a more complete sample of stars following Section 4.8 of Drlica-Wagner et al. (2021), 0 ≤ EXTENDED_CLASS_G ≤ 2, which has been found to be > 95% complete down to a magnitude of g ∼ 22.5 mag. We use this value as the g-band magnitude limit for our characterization analysis, paired with the i-band S/N = 10 magnitude limit of i ∼ 22.7 mag. We find that the derived properties of Eri IV are insensitive to moderate (±0.3 mag) variations in the assumed magnitude limits.

We searched for old, metal-poor stellar systems in the DELVE catalog using the simple ³² matched-filter algorithm, which has been applied to data from several DECam survey programs to help discover more than 20 Milky-Way satellites (Bechtol et al. 2015; Drlica-Wagner et al. 2015; Mau et al. 2019, 2020; Cerny et al. 2021). In brief, simple uses an isochrone matched-filter approach in color–magnitude space to increase the contrast of halo substructure relative to foreground Milky-Way stars at a given distance. Specifically, we split the DELVE catalog into HEALPix pixels (Górski et al. 2005) at nside = 32 (∼3.4 deg ²) and searched for stars consistent with a PARSEC (Bressan et al. 2012) isochrone of age τ = 12 Gyr and metallicity Z = 0.0001 ([Fe/H] ∼ − 2.2 dex). We scanned this isochrone over a range of distance moduli from 16.0 ≤ m − M ≤ 23.5 mag in steps of 0.5 mag. For each step in distance, we selected stars consistent with the isochrone according to , and then smoothed the resulting density field with a Gaussian kernel. Local spatial overdensities were identified by iteratively raising a density threshold until fewer than 10 disconnected peaks remained (Bechtol et al. 2015). The resulting candidates with significance > 5σ were visually inspected.

Eri IV was the most promising new candidate identified by the simple search. ³³ The corresponding density peak had a Poisson significance of 11.8σ relative to the stellar density in an annular background region and showed evidence of a prominent blue horizontal branch, lending significant credence to the detection. We note that this candidate system resides within the Pan-STARRS1 DR1 (PS1) footprint, but it was not identified in the census conducted by Drlica-Wagner et al. (2020). However, the observational selection function constructed by that work suggests that the detectability of a system with the properties of Eri IV is only ∼ 25% using the PS1 data, and thus the previous non-detection is not unexpected (see Section 3).

In Figure 1, we present diagnostic plots similar to those inspected during the search process. These include the smoothed stellar and galaxy density around the candidate, along with a color–magnitude diagram, and a (azimuthally averaged) radial profile centered at the location of the candidate system.

Zoom In Zoom Out Reset image size

We fit the morphological and stellar population parameters of Eri IV using the unbinned maximum-likelihood formalism implemented in the Ultra-faint Galaxy Likelihood (ugali) software toolkit (Bechtol et al. 2015; Drlica-Wagner et al. 2020). ³⁴ We modeled the 2D spatial distribution of stars with a Plummer profile and fit the system's color–magnitude diagram with a set of PARSEC (Bressan et al. 2012) isochrones assuming a Chabrier (2001) initial mass function. We simultaneously constrained the centroid (α₂₀₀₀, δ₂₀₀₀), semimajor axis (a_h), ellipticity (), and position angle (P.A.) of the Plummer profile and the age (τ), metallicity (Z), and distance modulus (m − M) of the isochrone by sampling the posterior probability distributions of each parameter using an affine-invariant Markov Chain Monte Carlo (MCMC) ensemble sampler (emcee; Foreman-Mackey et al. 2013). We then derived estimates for the system's azimuthally averaged angular and physical half-light radii (r_h and r_1/2, respectively), absolute visual magnitude (M_V ), average surface brightness within the half-light radius (μ), mean metallicity ([Fe/H]), and the total stellar mass of the system (M*).

We calculated the ugali membership probability for each star based on the Poisson probability that it belongs to Eri IV based upon its spatial position, flux, and photometric uncertainty, given an empirical model of the local stellar field population and of a putative dwarf galaxy. We also report the quantity ∑p_i,ugali , which can be interpreted as the number of Eri IV member stars brighter than the DELVE magnitude limits.

The resulting values for each parameter are shown in Table 1. We note that the uncertainties for the system's age and distance modulus are significantly asymmetric due to a bimodal posterior distribution. We believe that this behavior is likely introduced by foreground stars just brighter than the horizontal branch and do not attribute it to the presence of a second stellar population. We present the MCMC-derived posterior probability distributions for all parameters in Appendix.

Table 1. Morphological, Isochrone, and Proper-Motion Parameters for Eri IV

Parameter	Value	Units
α2000		deg
δ2000		deg
ah		arcmin
rh		arcmin
r1/2		pc
		...
P. A.		deg
m---M	a	mag
D⊙		kpc
τ		Gyr
Z	< 0.00013 b	...
∑i pi,ugali		...
TS	200.6	...
MV	−4.7 ± 0.2 c	mag
μ	28.2	mag arcsec−2
M*		M⊙
[Fe/H]	< − 2.08 b	dex
E(B − V)	0.107 d	mag
ℓ	209.499	deg
b	−27.772	deg
	+0.25 ± 0.06	mas yr−1
μδ	−0.10 ± 0.05	mas yr−1
∑i pi,MM	31.5 ± 3.7	...

Notes. The quoted uncertainties were derived from the highest density interval containing the peak and 68% of the marginalized posterior distribution.

^a We assume a systematic uncertainty of ±0.1 associated with isochrone modeling (Drlica-Wagner et al. 2015). ^b The metallicity posterior peaked at the lower bound of the allowed parameter range (Z = 0.0001); we therefore quote an upper limit at the 95% confidence level. ^c The uncertainty in M_V was calculated following Martin et al. (2008) and does not include uncertainty in the distance. ^d This E(B − V) value refers to the mean reddening of all sources within one half-light radius.

Download table as:  ASCIITypeset image

We utilized astrometric data from Gaia Early Data Release 3 (EDR3; Gaia Collaboration et al. 2021) to compute the systemic proper motion of Eri IV. We first filtered the sample by removing stars with non-zero parallax (ϖ − 3σ_ϖ > 0) and large proper motion (). We applied the following Gaia quality flags: ruwe < 1.4, astrometric_excess_noise_sig < 2, , duplicated_source=False, astrometric_params_solved > 3 (Lindegren et al. 2021; Riello et al. 2021). We removed background AGNs/QSOs that are cross matched to the Gaia catalog using gaiaedr3.agn_cross_id. We constructed a Gaussian mixture model composed of a satellite and Milky-Way background component fit to the proper motions and spatial positions of stars after applying a color–magnitude filter (Pace & Li 2019). We generally followed the methodology of Pace & Li (2019), but instead of fixing the spatial parameters of Eri IV, we assumed Gaussian priors based on the best-fit parameters from the ugali analysis as was done in Cerny et al. (2021).

We derived a systemic proper motion of and μ_δ = − 0.10 ± 0.05 mas yr⁻¹ and we found that the sum of the mixture-model membership probabilities is ∑_i p_i,MM = 31.5 ± 3.7. In the right panel of Figure 2, we present a proper-motion scatterplot of all stars meeting the cuts described above in gray and overlay the likely members identified by the mixture model in colors representing mixture-model-derived probabilities. In the center panel, we circle these stars in the DELVE color–magnitude diagram. The clear overdensity of stars with consistent proper motions further indicates that Eri IV is a co-moving system.

Zoom In Zoom Out Reset image size

In Figure 3, we show the best-fit parameters of Eri IV in the plane of absolute V-band magnitude and physical half-light radius (M_V − r_1/2) along with other known Milky-Way satellites, including classical globular clusters, faint halo star clusters, and candidate and confirmed dwarf galaxies (see Cerny et al. 2021, and references therein). The position of Eri IV in this space is consistent with the known population of ultra-faint dwarf galaxies, suggesting a dwarf galaxy classification for the system. In particular, the large physical size (r_1/2 = 75 pc) and ellipticity ( = 0.54) of Eri IV are inconsistent with the population of comparably bright globular clusters (Harris 1996, 2010 edition). Spectroscopic measurement of a large velocity dispersion for this system could provide evidence that the system is a dark-matter-dominated dwarf galaxy. Furthermore, spectroscopic measurements of a metallicity dispersion could provide independent evidence for the system's classification, since a large dispersion could imply that Eri IV has undergone multiple generations of star formation, which is more suggestive of a dwarf galaxy nature (e.g., Willman & Strader 2012).

Zoom In Zoom Out Reset image size

We also note that the predicted orbital properties of Eri IV based on sky position, distance, proper motion, and a range of possible line-of-sight velocities are slightly more consistent with the known population of dwarf galaxies than globular clusters (see Section 4.3), although the system's morphological properties provide stronger evidence for its classification.

The color–magnitude diagram in Figure 2 shows that the horizontal branch of Eri IV is well populated, with several stars residing in the instability strip, where Population II variable stars are likely to be found. In particular, based on the empirical relation from Martínez-Vázquez et al. (2019), a dwarf galaxy with M_V = − 4.7 mag is expected to contain 3 to 4 RR Lyrae (RRL) stars.

We searched the Gaia DR2 variability catalogs (Clementini et al. 2019; Holl et al. 2018) and found two RRL candidates within ( ∼ 6 r_h). However, the mean Gaia G-band magnitudes of these stars are ≳ 0.5 mag brighter than the horizontal branch of Eri IV measured in Gaia. The Gaia variability analysis reports a mean magnitude uncertainty of ∼ 0.002 mag with semi-amplitudes of ∼0.33 mag and ∼0.34 mag for these RRL candidates using 33 and 29 "clean" epochs of observation. The inconsistency between the mean magnitudes of these stars and the horizontal branch of Eri IV makes it unlikely that these stars are members. We examined the DELVE photometry for each of these RRL stars and confirmed that they are brighter than the horizontal branch of Eri IV by at least 0.3–0.5 mag. While the mean magnitudes of these RRL stars are inconsistent with the horizontal branch of Eri IV, the clustering of two distant halo RRL within such a small angular area is also unexpected. It is possible that these stars are incorrectly classified as RRLs. Another possibility is that these stars are being brightened by unresolved binary companions. More complete variability information from Gaia DR3 may provide additional insight into the nature of these stars.

We also searched the PS1 RRL catalog (Sesar et al. 2017) and found one additional candidate RRL star less than from Eri IV's centroid, but with an estimated distance modulus of 19.14 mag ( ∼ 0.3 mag closer than Eri IV). Inspection of the DELVE photometry for this star reveals that it is consistent with the isochrone locus shown in Figure 2, and was assigned a ugali membership probability of 0.98 and a mixture-model membership probability of 0.98 ³⁵ , although we note that the former depends on the mean magnitude of this object, which is sparsely sampled by DELVE. The Gaia G-band mean magnitude of this object, which is reported in Gaia EDR3 based on 22 epochs, is also consistent with the isochrone locus of Eri IV. Thus, it seems likely that this RRL candidate is a member of Eri IV, despite the inconsistent distance modulus reported by Sesar et al. (2017).

Lastly, we note that both the existing RRL catalogs are incomplete at faint magnitudes (Clementini et al. 2019; Sesar et al. 2017), and thus may not include all RRL members of this system. Given the sparse sampling of the DELVE data at the location of Eri IV, even when augmented by the existing PS1 data, we do not attempt to use it to identify RRL and instead note that follow-up observations are merited to identify and characterize variable stars in this system.

While the line-of-sight velocity of Eri IV is currently unknown, we can explore possible orbits by scanning over a range of line-of-sight velocities. We used the gala package to compute orbits with the MilkyWayPotential Galactic potential (Price-Whelan 2017), which consists of two Hernquist (1990) spheroids to model the bulge and nucleus, a Miyamoto & Nagai (1975) axisymmetric stellar disk, and an NFW dark-matter halo (Navarro et al. 1996). For a given heliocentric line-of-sight velocity, v_los, we computed the integrals of motion, E and L_z , and the orbital pericenter (r_peri), apocenter (r_apo), and eccentricity. For comparison, we computed the same properties for other Milky-Way satellite galaxies and globular clusters using the 6D phase-space information compiled in A. Pace et al. (2021, in preparation) and Vasiliev & Baumgardt (2021). In Figure 4, we show the orbital properties of Eri IV and the population of other known satellites. For values of − 180 ≤ v_los ≤ 420 km s⁻¹, we find that Eri IV is bound to the Milky Way.

Zoom In Zoom Out Reset image size

Interestingly, the observed Milky-Way satellite galaxies are preferentially found near their orbital pericenter whereas basic orbital dynamics predicts that satellites on elliptical orbits should spend more time near their apocenters (Fritz et al. 2018; Li et al. 2021). To compare Eri IV to the known population of satellite galaxies, we examined the ratio, f = (d_GC − r_peri)/(r_apo − r_peri), where d_GC is the distance to the Galactic center. The ratio f quantifies whether a satellite is close to its pericenter (f = 0) or apocenter (f = 1). As pointed out by Fritz et al. (2018), there is an observed excess of satellites with f < 0.1; in particular, 16/30 satellites with high-quality proper-motion measurements from Gaia DR2 were found to have f < 0.1, which they identify as a possible observational selection effect. In Figure 4, we find that Eri IV is only near its pericenter (f < 0.25) if it has a large eccentricity, which corresponds to the velocity extremes of the range we explored (v_los ≲ − 70 km s⁻¹ or v_los ≳ 315 km s⁻¹). Thus, Eri IV may be consistent with the prediction that many satellites will be discovered near their orbital apocenters (Fritz et al. 2018).

As seen in the leftmost panel of Figure 1, a clear overdensity of isochrone-selected stars is visible to the northeast of the densest region of Eri IV. This overdensity is also independently visible (albeit less prominent) in members selected solely on the basis of color and Gaia proper motion, supporting its characterization as a real structure. We note that despite the presence of Galactic cirrus, the E(B − V) measured by Schlegel et al. (1998) varies by ≲ 0.02 over the Eri IV field, making it unlikely that this feature is artificially introduced by spatial variations in the foreground dust.

Such extended stellar features can arise via tidal stripping of satellite member stars via interactions with the Milky Way (e.g., Küpper et al. 2012), but should be cautiously interpreted in systems like Eri IV where the total number of observed stars is low. While we defer more detailed spatial modeling and analysis of this candidate tidal feature until deeper photometric data are available, we investigated the possibility that our parameter estimates are biased due to the asymmetric spatial distribution produced by this overdensity. A similar situation was considered by Drlica-Wagner et al. (2015) for the tidally disrupting satellite Tucana III, and we followed their procedure by fitting an azimuthally symmetric ( = 0) stellar-density profile. We find that the azimuthally symmetric fit of Eri IV yields a consistent value of the azimuthally averaged half-light radius (r_h) within the quoted uncertainty. This suggests that the potential tidal feature is not significantly inflating the best-fit half-light radius of Eri IV (and thereby biasing the system's classification); however, we do observe that the best-fit centroid of Eri IV is slightly offset from the location of highest density (Figure 1; see also Figure 5).

We note that the proper-motion vector of the system is aligned nearly perpendicular to the potential tidal feature and the major axis of Eri IV (Figure 1). Whether or not this configuration favors or disfavors tidal disruption as the cause of the feature depends on the orientation of Eri IV's orbital plane and phase (e.g., Küpper et al. 2012).

Lastly, we note that our orbit model predicts that Eri IV has a pericenter of r_peri > 30 kpc, which is more distant than many other tidally disrupting systems (e.g., Palomar 5, Sagittarius, Tucana III). While hints of tidal disruption have been suggested for satellites with larger pericenter distances (e.g., Hercules, Leo V, and Canes Venatici II; Sand et al. 2012; Roderick et al. 2015), some of these claims have been called into question by deeper Hubble Space Telescope observations (Mutlu-Pakdil et al. 2019, 2020). Furthermore, Ji et al. (2021) recently emphasized the important gravitational influence of the LMC when modeling the orbit of the tidally disturbed system Antlia II. If Eri IV is indeed experiencing tidal disruption, it could provide valuable insight into the process of satellite disruption in the outer Milky-Way halo. Deeper photometry and radial-velocity measurements are critical to confirm this candidate tidal feature and to determine the full orbital characteristics of Eri IV.

While a number of recently discovered ultra-faint dwarf galaxies are likely to be presently associated with the Magellanic Clouds (e.g., Erkal & Belokurov 2020; Patel et al. 2020), this is unlikely for Eri IV based on its current location. Eri IV is located ∼ 68 kpc and ∼ 83 kpc from the LMC and SMC, respectively, and is not aligned with the Magellanic Stream. Nonetheless, our orbit model does not rule out the possibility of a past association between Eri IV and the LMC, and a more conclusive investigation will require radial-velocity information.

Eri IV is situated close in projection to the Sagittarius (Sgr) stream, leading us to investigate a possible association between the two systems. Inspecting the Sgr stream model from Vasiliev et al. (2021), we find that a wrap of the Sgr stream passes to the north of Eri IV at a heliocentric distance of ∼ 50 kpc. However, this is not especially well matched to the distance of Eri IV. Furthermore, we note that the measured proper motion of Eri IV is misaligned with the modeled proper motion of the Sgr stream at this location. Thus, we find no obvious connection between Eri IV and the Sgr stream. Likewise, we find that Eri IV is situated close in projection to the Gjöll stellar stream (e.g., Hansen et al. 2020), but note that Eri IV is > 40 kpc more distant than Gjöll's apocenter and has an inconsistent proper motion.

Lastly, it has been proposed that a substantial fraction of the observed satellite galaxies co-orbit the Milky Way in a thin plane known as the "Vast Polar Structure" (Pawlowski et al. 2012; Fritz et al. 2018). We find that the spatial position of Eri IV places it outside of the plane of this postulated structure.