Spectroscopy of Ultra-diffuse Galaxies in the Coma Cluster
TTo appear in Astrophysical Journal Letters
Preprint typeset using L A TEX style emulateapj v. 12/16/11
SPECTROSCOPY OF ULTRA-DIFFUSE GALAXIES IN THE COMA CLUSTER
Jennifer Kadowaki, Dennis Zaritsky, and R. L. Donnerstein
Steward Observatory and Department of Astronomy, University of Arizona, Tucson, AZ 85719
To appear in Astrophysical Journal Letters
ABSTRACTWe present spectra of 5 ultra-diffuse galaxies (UDGs) in the vicinity of the Coma Cluster obtainedwith the Multi-Object Double Spectrograph on the Large Binocular Telescope. We confirm 4 of theseas members of the cluster, quintupling the number of spectroscopically confirmed systems. Like thepreviously confirmed large (projected half light radius > > . (cid:46) − . Subject headings: galaxies: distances and redshifts — galaxies: general — galaxies: stellar content INTRODUCTION
Ultra-diffuse galaxies (UDGs) are a class of spatially-extended, low surface brightness galaxy. To understandtheir physical nature and study them in the context oftheir environment, we require accurate distance measure-ments.Spectroscopic redshifts have been measured for 4galaxies categorized as UDGs since the renewed interestin extreme low surface brightness galaxies in 2015. Of thepopulation discovered in proximity to the Coma Cluster,DF44 ( α = 13 h m . s ; δ = 26 ◦ (cid:48) (cid:48)(cid:48) ) is the only UDGspectroscopically confirmed as a Coma member (vanDokkum et al. 2015b). Likewise, a spectroscopic redshiftmeasurement places VCC 1287 ( α = 12 h m . s ; δ =13 ◦ (cid:48) . (cid:48)(cid:48) ) in the Virgo Cluster (Beasley et al. 2016)and DGSAT I ( α = 01 h m . s ; δ = 33 ◦ (cid:48) . (cid:48)(cid:48) )in a low density filament of the Pisces-Perseus Super-cluster (Mart´ınez-Delgado et al. 2016), identifies UGC2162 as the nearest UDG (Trujillo et al. 2017). WhileDF44 and VCC 1287 are projected near their associatedgalaxy clusters, other apparent close associations can bemisleading, as demonstrated by Merritt et al. (2016). Allcoordinates listed in this paper are given in J2000.In addition to providing the critical distance measure-ment, spectra provide information on the current stellarpopulations and star formation history of these galax-ies. For example, van Dokkum et al. (2015b) notedthat the spectrum of DF44 is similar to that of early-type galaxies, with Balmer and G-band absorption lines,concluding that DF44 is a quiescent galaxy with no sig-nificant on-going or recent star formation. Such find-ings inform the development of models for the originand evolution of these systems. We aim to determineif there does exist a class of physically extended UDGs,like DF44, in significant numbers in the cluster environ-ment and if these galaxies are exclusively quiescent, oldgalaxies. In § [email protected] analysis. In § m = 0 . Λ = 0 . H = 69 . − . DATA
We selected UDGs to observe from the van Dokkumet al. (2015a) sample of 47 candidates projected nearthe Coma Cluster. In particular, we selected luminous( M g (cid:46) − . r e , larger than 2.9 kpc assuming the UDG isin Coma) to focus on those analogous to DF44. Only18 UDGs from the parent sample satisfy both the sizeand luminosity criteria. Finally, we had to require thatpotential targets have a nearby guide star brighter than M R <
15 that is suitably located for the guide probe.Our final sample contained 10 candidate targets, of whichwe observed six. We list the observed UDGs and relevantparameters in Table 1.On 2016 March 9 and 10 we observed in mostly clearconditions using the Multi-Object Double Spectrograph(MODS; Pogge et al. 2010) on the Large Binocular Tele-scope (LBT) prior to the binocular mode being opera-tional. Because UDGs have relatively flat surface bright-ness profiles, we opted to use a wide slit to increase theamount of light collected, at the expense of spectral reso-lution. We used a custom 2.4 (cid:48)(cid:48) -wide slit, twice as wide asthe largest previously available long slit. We observedeach target for 60 to 90 minutes, combining multiple30 minute dithered exposures, simultaneously using theG400L grating (400 lines mm − blazed at 4 . ◦ centeredon 4000˚A with a resolution of 1850) in the blue channel(3200-5800 ˚A) and the G670L grating (250 lines mm − blazed at 4 . ◦ centered on 7600˚A with a resolution of2300) in the red channel (5800-10000 ˚A). We do not usethe red channel data due to higher sky brightness.We reduced the spectra using the MODS data reduc-tion pipeline (Pogge et al. 2010). Reduction includes thestandard steps of overscan correction, 2D bias subtrac-tion, dark current subtraction, flat fielding, wavelength a r X i v : . [ a s t r o - ph . GA ] M a r Kadowaki et al.
Fig. 1.—
Rest-frame, continuum-subtracted spectra of the ob-served UDGs for which we can measure a reliable redshift. Figure4 highlights the oxygen emission lines in DF08. We include the Aand K-type continuum-subtracted stellar spectra for comparison.We label the Balmer lines to guide the eye. calibration, sky subtraction, flux calibration, and extrac-tion. Because we use the MODS long-slit, over 80% ofthe slit is clear sky, allowing for high S/N sky subtrac-tion using the two-dimensional fitting available in thestandard pipeline. Our final spectra have a S/N in thecontinuum ranging from 3 to 8 per pixel. These are there-fore fairly low S/N spectra. We obtained these redshift-quality spectra for six targets, but exclude DF17 fromfurther discussion because ultimately we were unable tomeasure a redshift. RESULTS
We now describe our measurement of the individual re-cessional velocities of the UDGs and examine the natureof the spectra, both in terms of the individual spectraand a combined stack. We confine our discussion to theblue channel spectra (Figure 1). The most prominentabsorption features in the UDG spectra are the Balmerabsorption lines, the Ca II H & K lines just bluewardof restframe 4000 ˚A and the G-band at about 4300 ˚A.These features are also prominent in the published spec-trum of DF44 (van Dokkum et al. 2015b).
Redshifts and Coma Cluster Membership
We measure redshifts and uncertainties by cross-correlating the UDG spectra against spectral templates.The ubiquity of Balmer lines suggests that an A-typestellar spectrum would be an ideal template, althoughthe Ca lines and G-band suggest that later type starsmight also be suitable. As such, we use both, but gener-ally find more significant results using the A-type tem-plate, with velocities differing by no more than 60 kms − between templates in most cases, with one exceptionat 120 km s − .We use the IRAF task XCSAO for the cross- IRAF is distributed by the National Optical Astronomy Ob-servatory, which is operated by the Association of Universities forResearch in Astronomy (AURA) under a cooperative agreementwith the National Science Foundation. correlation analysis. We vary parameters, including thehigh and low frequency cutoffs, while searching veloc-ities between − − and 30 ,
000 km s − ) for thecorrelation function maximum. We only accept veloc-ities that are robust to changes in the input parame-ters and then visually confirm that multiple absorptionlines correspond to the principal absorption features thatwe mention above. We extract high confidence redshiftsfrom the spectra of 5 out of the 6 observed UDGs. Forthe sixth, we do not even extract a low confidence red-shift estimate.Because of heliocentric velocity corrections, differencesin slit illumination, and wavelength calibration errors,it is possible for template stars to have effective refer-ence velocities that differ by several tens of km s − fromthe published values. Such uncertainties are irrelevantwhen determining whether galaxies lie within the Comacluster, but the surprising availability of emission linesin one of our spectra allows us to correct the velocityzero point and also provides a test of the absorption linecross-correlation procedure. We will discuss this inter-esting object further below.We compare emission and absorption redshifts forDF08 to estimate the external uncertainties of the stel-lar cross correlation analysis. We find an offset of 220km s − between the recessional velocities measured fromcross-correlation with spectral templates and from theshift in [OII] and [OIII] lines with restframe wavelengthsin air of 3726.05, 3728.80, 4958.92, 5006.85 ˚A (about2 σ discrepant given the internal error estimate). Differ-ences among the cross-correlation results using differenttemplates for all of our UDGs are <
120 km s − . Weconclude that even if our uncertainties are truly as largeas 200 km s − , our conclusions regarding membership inComa for these UDG candidates is unchanged.We convert the spectra of the five UDGs from observedto rest frame, subtract the continua, and display themin Figure 1. The best-fit UDG recessional velocities anduncertainties are listed in Table 1.To determine Coma cluster membership, we place theUDGs on the cluster phase-space diagram, which we con-struct using measurements of galaxies retrieved from theNASA Extragalactic Database (Figure 2). There areenough known Coma members that they clearly delin-eate the classic wedge distribution defined by the causticcurves. Four of our UDGs (DF07, DF08, DF30, andDF40), plus the previously measured DF44, fall withinthe caustics and we conclude these are bona fide mem-bers. Another of our UDGs (DF03) lies well outside thecaustic curves and we conclude that this system is 45Mpc behind the cluster. From the statistics that 5 out ofthe 6 UDGs with measured redshifts lie within Coma andrestricting ourselves to systems that satisfy our selectioncriteria, we conclude that of the 18 van Dokkum et al.(2015a) systems that satisfy our size and magnitude cri-teria, ∼
15 are likely to be Coma cluster members. Weconclude that there is a significant population of physi-cally large cluster UDGs. We do not interpret the phase-space diagram further because the radial selection effectsin the original catalog are likely to be significant.
Stellar Populations and Metallicity pectroscopic Confirmation of Coma UDGs 3
TABLE 1Our Spectroscopic UDG sample a UDG RA Dec R b µ ( g, r eff M g t exp cz (J2000) (J2000) (arcmin) (mag arcsec − ) (kpc) (mag) (min) (km s − )DF03 c h m . s ◦ (cid:48) (cid:48)(cid:48) . +0 . − . . +1 . − . − . +0 . − .
90 10150 ± h m . s ◦ (cid:48) (cid:48)(cid:48) . +0 . − . . +1 . − . − . +0 . − .
90 6587 ± h m . s ◦ (cid:48) (cid:48)(cid:48) . +0 . − . . +1 . − . − . +0 . − .
60 7051 ± h m . s ◦ (cid:48) (cid:48)(cid:48) . +0 . − . . +1 . − . − . +0 . − .
90 ...DF30 12 h m . s ◦ (cid:48) (cid:48)(cid:48) . +0 . − . . +0 . − . − . +0 . − .
60 7316 ± h m . s ◦ (cid:48) (cid:48)(cid:48) . +0 . − . . +0 . − . − . +0 . − .
60 7792 ± a The coordinates, surface brightness, effective radii, and absolute magnitude are quoted from van Dokkum et al. (2015a). b The projected angular separation was computed from the center of the Coma Cluster ( α = 12 h m . s ; δ = 27 ◦ (cid:48) (cid:48)(cid:48) ). c Because DF03 is located behind the Coma Cluster, the values for the effective radius and absolute magnitude have been recomputedusing the new distance.
Projected Distance (Mpc) c z ( k m s ec ) r (arcmin) This workvan Dokkum , et al . (2015b) Fig. 2.—
Phase space diagram of the Coma cluster. The smallblack dots represent individual galaxies projected near the clus-ter, which mostly form the characteristic caustic pattern centeredon Coma’s mean recessional velocity (horizontal green line). Thedashed yellow line indicates Coma’s virial radius (Kubo et al. 2007).The large red star represents the location of DF44 (van Dokkum etal. 2015a), with uncertainties smaller than the symbol. The largeblue circles represent the locations of our UDGs. We concludethat four UDGs from our sample (DF07, DF08, DF30, DF40) andDF44 are cluster members and that DF03 is well behind the Comacluster.
The presence of Balmer lines in galaxy spectra is ofteninterpreted as evidence for intermediate age (1 Gyr) pop-ulations, but can also be the result of low metallicity. Wesuspect the latter is the case in these galaxies given theirred colors as a population (Koda et al. 2015). To explorethis issue further, we combine the spectra to produce ahigher signal-to-noise spectrum (Figure 3), but cautionthat this approach caries a risk that we are combiningsystems with different properties.We use the P´EGASE stellar population modeling soft-ware (Fioc & Rocca-Volmerange 1997) to create com-parison spectra for single age (13 Gyr) populations withthree different metallicities (Figure 3). We visually placethe stacked spectrum in the abundance sequence. In par-ticular, we focus on the strength of the H δ line and theratio of the H γ line to the G-band, which lies just to theblue of H γ . This analysis is clearly a preliminary deter- Fig. 3.—
A comparison of the UDG composite spectrum (secondfrom the top) and modeled stellar population spectra of differingmetallicity and age (continuum subtracted in all cases). The com-posite includes all five of our UDGs. We have placed it, on thebasis of visual inspection, in order among the metallicity sequenceon the basis of the H δ absorption depth and the ratio of the G-bandto H γ . The model spectra with [Fe/H] < − . mination of the typical metallicity of such UDGs, and acomposite one at that, but the indications are that theseare quite metal poor ([Fe/H] (cid:46) − . β and H γ +G-band, but results in an H & Kbreak that is too strong, as seen by the peak at around4000˚A, and a Mg and Fe complex at about 5200˚A that isnot in the observed spectra. Second, the low metallicityestimate for these galaxies agrees with a simple expec-tation from the metallicity-luminosity relation (Zaritskyet al. 1994, (ZKH)). For early type galaxies, g − B isroughly − . M g ∼ −
15 at M B ∼ − . − . Fig. 4.—
Smoothed spectrum of DF08, which shows [OII] and[OIII] emission lines at the recessional velocity derived from the ab-sorption lines. This is the first cluster UDG spectrum with emissionlines, demonstrating that some UDGs, even in the cluster environ-ment, retain some of their gas and host ionizing sources. Theapparent emission line labeled ‘?’ is interpreted as noise. It is ofcomparable magnitude to the apparent, but non physical, absorp-tion features near it. This level of noise is also responsible for thedeviation of the 4959/5007 line flux of 1:3.
Figure 13 of ZKH. Finally, independent measurementsof the metallicity in similar galaxies has also concludedthat these systems are of low metalliity (Makarov et al.2015). We conclude, based on the low metallicites, thatthese UDGs are not the tidal remnants of much largergalaxies.
Emission Lines
While absorption features are common in previ-ously observed UDG spectra (van Dokkum et al. 2016;Mart´ınez-Delgado et al. 2016), emission lines have beenfound only in one gas rich UDG (Trujillo et al. 2017).Surprisingly, we have a weak detection of oxygen in oneof the Coma UDGs in our sample. In Figure 4 we show aspectrum of DF08, which features weak [OII] and [OIII]emission lines at a redshift consistent with the absorp-tion line redshift. The existence of these lines indicatesthat DF08 is not entirely gas-depleted nor devoid of ion-izing sources. We are unable to localize a specific sourcein the 2D spectrum, so we conclude that the emission issomewhat physically extended. Emission lines may be-come a more common signature of UDGs as field samplesare explored (Di Cintio et al. 2017; Trujillo et al. 2017)and establishing the differences between field and clusterUDG samples is a clear next step. CONCLUSIONS
We present spectroscopy of five UDGs seen in projec-tion on the Coma Cluster with the MODS spectrographon the LBT and reach the following conclusions. • On the basis of their recessional velocities, we con-firm 4 of our UDGs to be Coma cluster members, thereby quintupling the population of spectroscop-ically confirmed Coma UDGs. • On the basis of its recessional velocity, we placeanother of our UDGs about 45 Mpc behind theComa cluster. Along with DGSAT I, DF03 is oneof the few field UDGs with a spectroscopic redshift. • Coupled with the spectroscopic confirmation ofDF44 (van Dokkum et al. 2015b), the result that5 of 6 spectroscopically observed, physically large(projected half light radius > ∼
15 of the 18similarly large Dragonfly UDGs are Coma mem-bers. • On the basis of a comparison between P´EGASEstellar population synthesis models and our com-posite UDG spectra, we conclude that, on aver-age, these systems are metal-poor ([Fe/H] (cid:46) − . • We present the first cluster UDG (DF08) spectrumwith emission lines. This finding demonstrates thatnot all cluster UDGs lack gas and sources of ioniz-ing radiation.The authors made use of the online cosmology calcu-lator as described in Wright (2006).The authors thank Barry Rothberg, our LBT supportastronomer, and Olga Kuhn who helped us to navigatethrough the MODS data reduction pipeline and to ac-quire the necessary blue and red spectral flats for ourcustom slit.The LBT is an international collaboration among in-stitutions in the United States, Italy and Germany.LBT Corporation partners are: The University of Ari-zona on behalf of the Arizona university system; Isti-tuto Nazionale di Astrofisica, Italy; LBT Beteiligungs-gesellschaft, Germany, representing the Max-Planck So-ciety, the Astrophysical Institute Potsdam, and Heidel-berg University; The Ohio State University, and TheResearch Corporation, on behalf of The University ofNotre Dame, University of Minnesota and University ofVirginia. This paper uses data taken with the MODSspectrographs built with funding from NSF grant AST-9987045 and the NSF Telescope System InstrumentationProgram (TSIP), with additional funds from the OhioBoard of Regents and the Ohio State University Officeof Research. The NASA/IPAC Extragalactic Database(NED) is operated by the Jet Propulsion Laboratory,California Institute of Technology, under contract withthe National Aeronautics and Space Administration.
Facilities:
LBT (MODS).
REFERENCESBeasley, M. A., Romanowsky, A. J., Pota, V., Navarro, I. M.,,Martinez Delgado, D. Neyer, F. & Deich, A. L., 2016a, ApJ,819, 20 Di Cintio, A., Brook, C. B., Dutton, A. A., Macci`o, A. V.,Obreja, A., & Dekel, A., 2017, MNRAS, 466, 1L pectroscopic Confirmation of Coma UDGs 5pectroscopic Confirmation of Coma UDGs 5