DISCOVERY OF AN ULTRA-DIFFUSE GALAXY IN THE PISCES-PERSEUS SUPERCLUSTER

The first data set was collected with a 15 cm aperture f/7.3 Takahashi TOA-150 refractor telescope located at the Fosca Nit Observatory (FNO) at the Montsec Astronomical Park (Ager, Spain). We used a STL−11000 M CCD camera with a large FOV (19 × 13, 169 pixel⁻¹). A set of individual 1200 s (luminance-filter) images were obtained in photometric nights between 2012 August and September, with a total integration of 43,800 s. Each sub-exposure was reduced following standard procedures for dark subtraction, bias correction, and flat fielding.

Deeper imaging of the field was acquired with a 0.4 m aperture, f/3.75 corrected Newton telescope located at the Remote Observatory Southern Alps (ROSA, Verclause, France) in 2013 August. We used a FLI ML16803 CCD detector, which provides a total field of 81' × 81' with a scale of 1237 pixel⁻¹. The total exposure time in the luminance filter was 13,200 s and the images were reduced using the same standard procedure.

Both images obtained to confirm the discovery of DGSAT I are shown in Figure 1. The luminance images taken with these amateur telescopes were first astrometrically calibrated (Lang et al. 2010), and then calibrated to the SDSS-DR7 (Abazajian et al. 2009) photometric system. To begin the flux calibration of both images, we remove any residual large-scale sky gradients in our wide-field images by modeling the background using a two-dimensional, fourth-order Legendre polynomial that was fitted to the median flux in coarse pixel bins after masking all sources that were detected above the background at ≳5σ level. The bins were typically 5' on a side.

Zoom In Zoom Out Reset image size

We then located 67 isolated stars distributed throughout each image with SDSS r-band magnitude 16 < r < 19, and performed aperture photometry on them. The same set of stars was measured in both images. The positions of these stars were matched with the SDSS DR7 catalog, from which we retrieved their g, r and i magnitudes. From these data, a linear relation between the r magnitude and the luminance instrumental magnitude was derived individually for each image. We find that the residuals from this relation are a strong linear function of g − i color and vary by 0.5 mag over the color range 0.4 < g−i < 2.7. Once this color term was corrected, the statistical uncertainty in the flux calibration is only 0.04 mag. The 5σ limiting surface brightness for background variations measured following the procedure of Cooper et al. (2011; see their Appendix) is 27.16 mag arcsec⁻² and 26.03 mag arcsec⁻² for the TOA-150 and 0.4 m f/3 data, respectively, as determined by measuring the standard deviation of the median of random 20'' apertures placed around each image. This means that the TOA-150 imaging is ∼1 mag deeper than the ROSA 0.4 m data in terms of photon statistics, and its sensitivity to large-scale surface brightness variations is less hindered due to the better quality of the flatfielding.

We use a deep image of the field around the And II dSph obtained with the Subaru/SuprimeCam wide field imager (34' × 27' FOV, 0202 pixel; Miyazaki et al. 2002), which is publicly available from the SMOKA archive (Baba et al. 2002). These observations include dithered exposures of 5 × 440 s in the Subaru V-band and 20 × 240 s in the Subaru I-band with a seeing around 06 (McConnachie et al. 2007; Ho et al. 2012).

Preprocessing of the data was done by debiasing, trimming, flat fielding, and gain correcting each individual exposure chip-by-chip using median stacks of nightly sky flats. The presence of scattered light due to bright stars both in and out of the field of view required removing this smoothly varying component before performing photometry and solving for a World Coordinate System solution. To remove scattered light, we fitted the smoothly varying component by creating a flat for every chip within each frame by performing a running median with a box size of 300 pixels. This was then subtracted from the original, unsmoothed frame to -produce a final image for photometric processing. The resulting Subaru V-band image is given in Figure 2.

Zoom In Zoom Out Reset image size

Photometric calibration was performed using the daophot ii and allstar (Stetson 1987) packages. Sources were detected in the V-band image by requiring a 3σ-excess above the local background, with the list of detections being subsequently used for the allstar run on the I-band image. To obtain clean CMDs and minimize the contamination by faint background galaxies, sources with the ratio estimator of the pixel-to-pixel scatter χ < 2.0 and sharpness parameters $| S| \lt 1.0$ (see Stetson 1987) were kept for further analysis. Instrumental magnitudes were obtained by cross-correlating with stellar photometry in the Subaru filter system from McConnachie et al. (2007), kindly provided in digital form, with zero-points determined as resistant mean of the differences between our instrumental and the calibrated Subaru magnitudes (denoted here as V', I'). The accuracy (standard error) of the zero-point is better than 0.03 mag. For the transformation to the standard system we combined Equation (1) in McConnachie et al. (2007), which link the Subaru and the INT filter systems, with the transformation equations for the INT and Johnson/Kron-Cousins systems, available on the INT Wide Field Survey (WFS) webpage.²⁰ The resulting transformations relations are:

$$\begin{eqnarray*}V & = & V^{\prime} +0.040\times (V^{\prime} -I^{\prime} )\\ I & = & I^{\prime} -0.090\times (V^{\prime} -I^{\prime} ),\end{eqnarray*}$$

where (V, I) and (V', I') are the magnitudes in the standard and Subaru systems, respectively.

During the refereeing process of this paper, spectroscopic observations of DGSAT I were obtained using the primary focus of the 6 m Bolshoi Teleskop Alt-azimutalnyi (BTA) telescope of the Special Astrophysical Observatory of the Russian Academy of Sciences (SAO RAS) with the SCORPIO spectrograph²¹ (Afanasiev & Moiseev 2005) on 2014 October 28th–29th. The slit width was 1'' and its length was 6'. The instrumental setup included the CCD detector EEV 42–40 with a pixel scale of 0.18 arcmin pixel⁻¹ and the grism VPHG1200B (1200 lines mm⁻¹) with a resolution of FWHM ∼ 5 Å, a reciprocal dispersion of ∼0.9 Å pixel⁻¹ and a spectral range of 3700–5500 Å. Two spectra of 12,000 and 13,200 s exposure time were obtained in two consecutive nights under very good atmospheric conditions. At the beginning and at the end of each night a sequence of 10–20 bias frames was recorded. The fluctuations of the bias level were small (∼4 e⁻). Flatfield and He–Ne–Ar arc calibrations were taken each science exposure, which allowed us to provide a reliable wavelength calibration and a pixel-by-pixel sensitivity correction for our spectra.

The data reduction was performed using the European Southern Observatory Munich Image Data Analysis System (MIDAS; Banse et al. 1983) and the Image Reduction and Analysis Facility (IRAF) software system (Tody 1993). It included bias and dark subtraction, flat-field correction, wavelength calibration, background subtraction and extraction of one-dimensional spectra. The dispersion solution provided an accuracy of the wavelength calibration of ∼0.16 Å. Because our object is of very low surface brightness, accurate subtraction of night sky lines is critical for the correct determination of its radial velocity. The sky background subtraction and correction of the spectrum curvature were done using the background task in IRAF. Finally, one-dimensional spectra were extracted from the two-dimensional ones and summed. The signal-to-noise ratio (S/N) in the integrated light spectrum of DGSAT I is ∼5 per wavelength bin and per spatial resolution element. The resulting integrated spectrum is displayed in Figure 3.

Zoom In Zoom Out Reset image size

Narrow-band observations were made with the 6 m telescope at the SAO RAS using the SCORPIO detector with a 2048 × 2048 pixel matrix in a 2 × 2 binning mode and an image scale of 018 pixel⁻¹, which yields a full field of view of 61 × 61. Images in Hα+[Nii] and in the continuum were obtained on 2013 October 26, by observing the galaxies through a narrow band Hα filter (FWHM = 75 Å) with an effective wavelength of 6555 Å, and the SED607 (with FWHM = 167 Å, λ = 6063 Å) and SED707 (with FWHM = 207 Å, λ = 7036 Å) intermediate band filters for the continuum. The exposure times were 6 × 300 s in the continuum and 3 × 600 s in Hα.

A standard procedure for analysis of direct CCD images was used for processing the data. The bias was initially subtracted from all the data, and then all the images were divided by a flat field. Afterwards, the cosmic ray events were removed and the sky background was measured and subtracted from each image. The continuum images for each pointing were normalized to the corresponding Hα images using 15 field stars so that continuum could be subtracted from the Hα images. The Hα flux was then measured from the continuum subtracted images and calibrated using spectrophotometric exposures of standard stars obtained on the same night. For DGSAT I we detected no Hα emission, with an upper limit for its Hα flux of log F(Hα)(ergs⁻¹ cm⁻²) < −15.8. This limit will be used to set constraints on the stellar populations in Section 3.4.

Figure 4 shows the CMD of a selected region of 815 × 1503 from the Subaru field (corrected for Galactic extinction A_B = 0.27) with RGB stars from the halo of M31 and some Galactic foreground stars at (V − I)₀  ≥  2.0. Based on the densely populated CMDs from the PAndAS survey (McConnachie et al. 2009) of M31's halo (e.g., Figure 4 in Conn et al. 2011), we selected stars with (V − I)₀ ≤ 2.0 for our sample, eliminating most of the Galactic foreground population but still keeping most of the stars in M31's halo and potential resolved stellar sources in the dwarf galaxy. Using a similar color criterion as Conn et al. (2012) to select putative RGB stars in the satellite candidate (Figure 4), we find that only 17 stars fall in this area of the CMD and within 40'' of the center of the galaxy, consistent with the average density of stars within these color cuts in the image. Hence, no significant over-density of resolved RGB stars is found at the location of DGSAT I that would indicate it is a dwarf satellite of M31. This is consistent with the lack of detection of this object in the list of over-density candidates in this region of Andromeda from the PAndAs survey (Martin et al. 2013). Thus, we conclude that the galaxy is not resolved into stars. This further rejects the possibility that DGSAT I is a companion of the more distant spiral galaxy NGC 404 that is 275 distant in projection (∼3.13 Mpc; Williams et al. 2010). It is also very unlikely that it is a dwarf member of any other galaxy group in the background of M31 in the local universe (e.g., NGC 672 with five known companions or NGC 891 with 18 known companions).

Zoom In Zoom Out Reset image size

To derive radial velocities of our object we used: (a) the fxcor task in IRAF; and (b) the ULySS program (Koleva et al. 2008, 2009) with Vazdekis et al.'s (2010) SSP model, the Salpeter IMF (Salpeter 1955) and the MILES stellar library (Sanchez-Blazquez et al. 2006). The task fxcor uses the Fourier cross-correlation method developed by Tonry & Davis (1979). The resulting radial velocity is V_h = 5453 ± 111 km s⁻¹. The cross-correlation peak is weak and the velocity error is large because of the low S/N in the spectrum.

The ULySS program provides more accurate results than fxcor, because it allows one to take into account the line-spread function (LSF) of the spectrograph, viz. the variation of the measured velocity and instrumental velocity dispersion as a function of wavelength (Koleva et al. 2008). The LSF was approximated by comparing a model spectrum of the Sun to twilight spectra taken during the same night as the studied object. The result is V_h = 5450 ± 40 km s⁻¹. Using the prescriptions of Tully et al. (2008), we transform our initial measured velocity to one relative to the motion of the Local Group and obtain V_LG = 5718 ± 40 km s⁻¹. This radial velocity corresponds to the Hubble distance of 78 ± 1 Mpc for H₀ = 73 km s⁻¹ Mpc⁻¹. The estimated velocity dispersion in the galaxy is σ_V ∼ 100 km s⁻¹. The σ_V value is very approximate given the low S/N in the spectrum of DGSAT I and needs to be verified at higher resolution and at higher S/N.

Interestingly, this velocity is consistent with those of the Pisces-Perseus supercluster, which has typical velocities in the range of 4800–5400 km s⁻¹. Figure 5 shows the position of DGSAT I, clearly overlapping a prominent filament of this supercluster in this sky region (Wegner et al. 1993) and in proximity to some of the massive members of this structure (NGC 383 and NGC 507, at projected distances of 2–3 Mpc). Therefore DGSAT I could well be part of this structure, lending support to our redshift distance estimate, with a corresponding diameter (effective size) near ∼10 kpc (see the next section). Its simultaneous low surface brightness (∼25 mag arcsec⁻²) and large physical extent would place it in the category of the ultra-diffuse galaxies (see Section 4).

Zoom In Zoom Out Reset image size

The structural parameters of the dwarf galaxy were determined by running galfit3 (Peng et al. 2010) on the sky-subtracted Subaru V- and I-band images, with foreground and background objects removed. We characterize the photometric properties of the galaxy by using galfit3 to fit a two-dimensional Sérsic profile,

$$\begin{eqnarray*}I(R) & = & {I}_{c}\mathrm{exp}[-{b}_{n}{(R/{R}_{e})}^{1/n}]\\ {I}_{c}\;=\;I(0) & = & {F}_{\mathrm{tot}}\;{[2\pi {qn}{({R}_{e}{b}_{n}^{-n})}^{2}{\rm{\Gamma }}(2n)]}^{-1},\end{eqnarray*}$$

where I is the surface brightness within ellipses with semimajor axes R, and Γ is the Gamma function. The conversion to the magnitude system (see Table 1) is given by μ(R) = ZP − 2.5logI(R), where ZP is the magnitude zero-point and μ_c = μ(0) the central surface brightness. The total flux F_tot (expressed as magnitude, m = ZP − 2.5logF_tot), the effective radius R_e, Sérsic index n, and axis ratio q = b/a are adjustable parameters. The constant b_n is determined by the requirement $F(\lt {R}_{e})\propto {\int }_{0}^{{R}_{e}}I(R)R\;{dR}\;=\;{F}_{\mathrm{tot}}/2$. The geometry of the light distribution is additionally characterized by the location of the profile center and the position angle of the semimajor axis.

We perform masking of intervening objects (mostly foreground stars) and background subtraction iteratively with SExtractor (Bertin & Arnouts 1996) in the area where we fit our model (280'' × 280''), and add user-defined masked regions around the brightest stars. In order to obtain the background reliably, we also mask the target galaxy generously "by hand." After a first pass of simultaneous object detection and background estimation, the task is repeated with the previously detected objects masked, which gives us a more accurate local background level and background noise, and thus also a more complete object mask. Finally, by reapplying SExtractor on the galfit residual image, we eventually also mask previously undetected objects that overlap with our target.

The noise map, which enters the χ² computation during fitting, was constructed by measuring the noise in the background and adding in quadrature the Poisson noise of the sources. In lieu of including the sky background as a free parameter in the galfit model, which can induce degeneracy in the output model parameters, we subtract the background from the image before fitting. Lastly, to account for the effect of seeing and telescope optics, we construct a model image of the PSF using several bright, unsaturated foreground stars near the position of DGSAT I in the image. This empirically derived PSF model is convolved with the galfit model before comparison to the image data.

We first fit a 2D-Sérsic profile as a basic model of the V- and I-light distribution of DGSAT Iİn this simple model, we find DGSAT I is almost circular (b/a ∼ 0.9) with a shallow, sub-exponential inner profile (n < 1), both of which are typical structural characteristics for galaxies of dwarf spheroidal morphology. However, despite the overall agreement of the model with the data (see Figure 6), the data show two peculiarities with respect to the smooth regular Sérsic model.

Zoom In Zoom Out Reset image size

First, there is a faint, fuzzy irregularly shaped elongated over-density of ∼6'' length near, but not coinciding with, the galaxy center (offset by ∼1''). Apart from its shape, the over-density also differs from the main body of the galaxy by its bluer color, more specifically V − I = 0.7 for the over-density as compared to V − I = 1.0 for a regions at a similar radius within DGSAT I$\dot{{\rm{D}}}$ue to its small angular size and the spatial resolution of our ground-based imaging, it is not possible to shed additional light on the nature of the blue over-density at this time (see also Section 3.4). It could be an unresolved clump of young stars overlapping the smooth (and possibly older) component of the galaxy, similar to those observed in some nearby dIrr galaxies (e.g., Sextans A, that contains a conspicuous, off-center star formation region; Dohm-Palmer et al. 1997). Alternatively, DGSAT I could be a nucleated UDGs, similar to those found in the Virgo cluster (Beasley et al. 2016; Mihos et al. 2015).

For this reason, we construct three galfit3 models that differ in how this blue off-center over-density is treated. Model 1 assumes that it should be included in the fit, and leaves it unmasked. Model 2 also leaves it unmasked, but accounts for it as an additional structural component with fully independent parameters from the main body of the galaxy (i.e., both the general parameters and those specific to the Sérsicf¸it). In Model 3 we exclude the over-density from the fit by masking it. Table 1 compares the resulting parameters of the main component (the sole component in Model 1 and Model 3). Parameters generally vary only mildly between these models, by ∼0.01 for m, R_e and q = b/a. Differences in the best-fit μ_c and n are more pronounced (∼0.1) as they more specifically reflect the central profile.

Another peculiarity in the light distribution of DGSAT is its significant lopsidedness: the center of the innermost isophotes are offset from those of the outermost isophotes by several arcseconds to the north–west. We accommodate for the apparent lopsidedness by allowing a first-order Fourier mode in the model isophotes. The best-fit amplitude of the mode is 0.21 (0.15) in V-band (I-band). Although allowing for the Fourier mode in a single-component system is not essential to obtain accurate flux and scale parameters, this modification to the model is useful for us since it provides a measurement of the galaxy center and a realistic set of isophotes, which we utilize to measure the surface brightness profile μ(R) with the IRAF task ellipse (see below).

When measuring μ(R) on the data image, we use the isophotes of the model. Fitting the isophote geometry simultaneously with μ(R) (which was done on the model image) is impeded at small radii due to the low central surface brightness, the extremely shallow central gradient, and the irregularities introduced by the off-center over-density and remaining small contaminants.

For the following discussion of the DGSAT I properties, we select Model 3 (where the over-density is masked) as our adopted model. We do not base this selection on the value of χ² at the best-fit parameters, since it is very similar for all three models considered.²² We assume that, given its different color and presumably younger population, this over-density is a peculiar feature of DGSAT I and its light is not reflective of the global properties (including the stellar mass) at same confidence as as in the smooth, red component of the galaxy. A rough estimate of the flux and size of this over-density can be obtained from the extra-component fitted in Model 2, yielding 2.5% flux fraction in the V-band (1.1% in the I-band), and 22% of the effective radius (both bands; see Table 2).

Table 2.  Summary of DGSAT I Properties, Assuming our Redshift Distance of 78 Mpc

Quantity	Notation	Value
Right Ascension	R.A.	01h 17m 3559
Declination	decl.	+ 33°31'4237
Radial velocity (heliocentric)	Vh	(5450 ± 40) km s−1
Hubble distance	D	(78 ± 1) Mpc
Apparent magnitude	mV	18.18 ± 0.04
	mI	17.17 ± 0.05
Central surface brightness	μc,V	(24.8 ± 0.2) mag arcsec−2
	μc,I	(24.0 ± 0.2) mag arcsec−2
Absolute magnitude	MV	−16.3 ± 0.1
	MI	−17.3 ± 0.1
Luminosity	LV	(2.7 ± 0.2) × 108 L⊙,V
	LI	(3.6 ± 0.2) × 108 L⊙,I
Total color	V − I	1.0 ± 0.1
Central color	(V − I)c	0.8 ± 0.1 (0.71 ± 0.02)
Effective radius	Re, I	(4.7 ± 0.2) kpc
Axis ratio	qI = (b/a)I	0.87 ± 0.01
Sérsic index	nI	0.6 ± 0.1
Over-density luminosity		∼7 × 106 L⊙,V (∼2.5% of DGSAT I)
		∼4 × 106 L⊙,I (∼1.1% of DGSAT I)
Over-density effective radius		∼1.0 kpc (both bands, ∼22% of DGSAT I)
Stellar mass-to-light ratio in I-band	M⋆/LI	1.1 M⊙/L⊙,I
Stellar mass	M⋆	4.0 × 108 M⊙
Gas mass (HI)	log MHI(M⊙)	<8.8
Hαflux	log F(Hα)(ergs−1 cm−2)	<−15.8
Star formation rate	log SFR (M⊙ yr−1)	<−2.6
Specific star formation rate	log sSFR (yr−1)	<−11

Note. The extinction-corrected apparent magnitudes in the standard (Johnson-Cousins) V- and I-band, central surface brightness, luminosity, axis ratio and size are the best-fit galfit3 2D-Sérsic profile constrained by our subaru images (see Section 3.3). R_e, b/a and n are given for the I-band, but differ from V only at the percent level. For the central color, two values are given: the central magnitude difference of the Sérsic models in the two bands, and (in brackets) as measured in a small 14 × 06 elliptical aperture. The photometric errors give the approximate systematic uncertainty from the choice of parametric model, except for the aperture-based central color where it is based on a Monte-Carlo realization of the aperture geometry. The error in distance and the calibration error of 0.03 mag was added in quadrature where applicable, while random errors from pixel noise are negligible and ignored. Due to the unknown systematic uncertainty in M_⋆/L_I, no errors are given here and for the derived M_⋆.

Download table as:  ASCIITypeset image

The apparent galaxy parameters (Table 1, column for "Model 3") translate to physical parameters using our adopted redshift distance of 78 Mpc. The extinction-corrected apparent I-band magnitude of 17.17 implies an absolute magnitude, M_I = −17.29, and, adopting M_⊙,I = 4.1 (Binney & Merrifield 1998), a luminosity L_I = 3.6 × 10⁸L_⊙,I. In the V-band, we obtain m_V = 18.18, which implies an absolute magnitude, M_V = −16.28 and, with M_⊙,V = 4.8, L_V = 2.7 × 10⁸ L_⊙,V. At the same time, the effective radius of 125 translates to 4.7 kpc. There is some uncertainty in R_e, both due to the variation with band as well as due to background uncertainty; we give here the maximum range of R_e resulting from fitting different models and both bands (see above and Table 1), which is (131–113)/2 = 09, hence ≈10% or ≈0.5 kpc maximum range in R_e. Thus, this galaxy is very extended (about Milky-Way sized; Bovy & Rix 2013) for its luminosity (which is less than 1/100 of the Milky Way).

In order to estimate the stellar mass of DGSAT I we use the work of Zibetti et al. (2010), who provide mass-to-light ratios (M_⋆/L) in SDSS bands for galaxy stellar populations as a function of one or more colors. They also show that the mass-to-light ratio in the i-band is well determined (low scatter and low dust effects) by one color alone, g − i. Fortunately, our Subaru images provide photometry in closely related bands, however we still need to transform the Johnson-Cousins system to the SDSS system.

With only one color at hand, we have little knowledge of the stellar population parameters (primarily age and metallicity) that effect the transformation between photometric bands. We refer to Jordi et al. (2006) when calculating g − i = 1.481 × (V − I) − 0.536 = 0.960 with our measured V − I = 1.01; this transformation is based on Stetson's extension of the Landolt standard stars and hence a mix of stellar populations. A transformation from Johnson I to SDSS i is unfortunately not possible from the tables provided in Jordi et al. (2006); we rely on the relations of R. Lupton, which are available on the SDSS website²³ and use the same set of standard stars. We obtain m_i ≡ i = I + 0.127 × (V − I) + 0.320 = 17.62, corresponding to log(L_i/L_⊙,i) = 8.57 when adopting an absolute solar magnitude of M_⊙,i = 4.58 (Blanton et al. 2003). With $\mathrm{log}[({M}_{\star }/{L}_{i})\;/\;({M}_{\odot }/{L}_{\odot ,i})]\;=\;0.963\quad +\quad 1.032\times (g-i)$, we obtain a stellar mass-to-light ratio M_⋆/L_i = 1.07 M_⊙/L_⊙,i and a stellar mass for DGSAT I of M_⋆ = 4.0 × 10⁸M_⊙.

We mention that the stellar M/L is an uncertain quantity; among other factors it is sensitive to priors in the star formation history and assumptions on the shape of the stellar initial mass function. For example, the relations of Bell et al. (2003) yield M_⋆/L_i = 2.2 M_⊙/L_⊙,i for the same g − i color; a factor of two higher than the result based on Zibetti et al. For comparison, using either only Population-I stars or Population-II stars, g − i is 0.97 or 0.93, i.e., the stellar population effects the color transformation by ∼0.04 mag, and thus M/L_i by a factor of ≲0.04. The i-band magnitude is even less sensitive to stellar population differences than the g − i color. For example, using the SDSS-Johnson transformation derived for Population-I stars from Jordi et al., in conjunction with R − I colors from Caldwell et al. (1993), m_i differs by a mere 0.002 mag from our adopted transformation using all standard stars.

However, the relatively large systematic uncertainty in M_⋆/L does not effect the conclusions of our study: that DGSAT I is extremely extended for its mass (or luminosity) and therefore may represent a possible class of peculiar galaxies with an extremely low surface brightness (that is typical of the much smaller dSph galaxies) for their mass (or size). We therefore attempt to explore further properties of DGSAT-I from the available data, namely Hα flux, SFR, and HI content, that may help to shed light on its formation.

The lack of spatially extended emission lines in the SAO spectra (see Section 2.3) is consistent with our Hα-band observations and rules out the possibility of this object being a nearby star-forming galaxy. This is also consistent with the observed upper limit of Hα flux detected in our SAO narrow band observations (Section 2.4). We also note that DGSAT I was not detected in the 21 cm survey of this filament of the Pisces-Perseus superclusters by Giovanelli & Haynes (1989). Taking the upper limit for the flux of this survey, F(HI) = 0.5 Jy km s⁻¹(with signal-to-noise S/N ∼ 5), this yields log M_HI(M_⊙) < 8.76.²⁴ We hence estimate the logarithm of the ratio of HI-to-total stellar masses, log M_HI/M_⋆ < 0.16. Accepting this upper limit, implies about as much mass in HI as in stars. This is, for example, four times HI fraction than that found in the LMC (Kim et al. 1998), which has one of the highest HI fractions of the galaxies in the local universe. Therefore we conclude that the upper-limit estimate is a very conservative one, and presume that M_HI of DGSAT I is actually at least one magnitude lower.

Following Kennicutt (1998), we determined the integral SFR in DGSAT I by the relation,

$$\begin{eqnarray}&&\mathrm{log}\;\mathrm{SFR}({M}_{\odot }\;{{\rm{yr}}}^{-1})=8.98+2\mathrm{log}(D[{\rm{Mpc}}])+\mathrm{log}({F}_{c}({\rm{H}}\alpha ))\end{eqnarray} \tag{ 1 }$$

where F_c(Hα) is its integral flux F in the Hα line in erg cm⁻² s (see Section 2.4), corrected for the Galactic extinction A(Hα) = 0.538A_B (Schlegel et al. 1998, with A_B = 0.27). The internal extinction in the dwarf galaxy itself was considered negligible. Therefore, for the distance of DGSAT I, we obtain an upper limit of log(SFR(M_⊙ yr⁻¹)) < −2.56. The corresponding upper limit for the specific star formation rate (sSFR; the SFR per unit galaxy stellar mass) is sSFR < −11.16.

Karachentsev & Kaisina (2013) discussed star formation properties of Local Volume galaxies. The SFR of DGSAT I is somewhat low for its upper limit HI mass, but still consistent with the spread in their sample (see their Figure 5). They showed that the median value of the sSFR does not change significantly for local volume galaxies with neutral hydrogen masses in the range 7 < log M_HI (M_⊙) < 9.5. Most galaxies have log sSFR(yr⁻¹) ∼ −10, but with a large scatter toward the lower sSFR over the whole range of M_HI. Some objects with neutral hydrogen masses in the range 7 < log M_HI [M_⊙] < 9.5 reach logarithmic sSFRs as low as ∼−15. Thus, the upper-limit sSFR and SFR/M_HI ratio of DGSAT I is consistent with their Local Volume study, and comparable with the values typical for quenched galaxies. However, SFR or hydrogen mass may also be much lower, a possibility that will have to await additional data to be tested.

DGSAT I displays a clear off-center over-density discussed in Section 3.3 (clearly visible in Figures 2 and 7). This raises the question of whether this feature harbors a young stellar population, originating in its last episode of star formation. Interestingly, irregardless of the our treatment of the blue over-density in our galfit3 modeling (Table 1), the central V − I color of the Sérsic profile is still ∼0.2 mag bluer than its global mean color (compare the V − I rows in the top two sections, m and μ_c, in Table 1). It is not clear if this difference is real or an artifact from profile mismatch or background uncertainties. Thus, we perform an additional test by comparing the mean color in two different regions: (1) a circular annulus between 4'' < R < 15'' from the center of DGSAT I and (2) a small elliptical aperture at the center with semimajor and minor axes a = 14, b = 06 with a position angle of PA = 10° north–west. These two annuli are indicated in the left panel of Figure 7(and described in the caption). As can be seen in Figure 7, these two regions are largely selected based on the features of the V − I color map, with the region (1) selected to both average over background fluctuations but also to avoid contamination from the blue over-density and region (2) selected to sample the core of the object. From this analysis, we find for region (1) V − I = 1.00 and for region (2) V − I = 0.71, which broadly confirms the bluer nature of the center as measured in the galfit3 modeling (Table 1). To ascertain the significance of this finding and evaluate any subjectivity in the definition of the regions, we perform a Monte-Carlo re-analysis with N = 400 samples, where the (i) center and (ii) specific parameters for the region shapes are allowed to vary. More specifically, for region (1), the inner and outer radius and for region (2), the semimajor and semiminor axes (a, b) as well as the position angle. The center is allowed to shift by 20% of the harmonic axis mean (0.2 $\sqrt{{ab}}$), the direction of the shift is random, R_in, R_out, a and b are varied by ±20% of their given value, and PA by ±10°. All variations are independent of one another and uniformly distributed within their bounds. As a result, the geometric color uncertainty is small, with a standard deviation of 0.01(0.02) mag in the outer (central) aperture, which confirms that the bluer color for the center is robust.

Zoom In Zoom Out Reset image size

It is difficult to assess with the current data what this 0.3 mag color difference in the center implies about a possible stellar population gradient. The color offset, location and shape of the over-density is similar to those reported for dwarf elliptical (dE) galaxies by Lisker et al. (2006), which were interpreted as the presence of young stars overlaying the mass-dominant old population. It is therefore likely that the bluer color of the over-density feature of DGSAT I (V − I ∼ 0.7) indicates the presence of young stars from a recent episode of star formation in this region of the galaxy.