Engineering a New Way to Extend Star Spectroscopy
A combination of computer science + astrophysics techniques.
Wednesday, January 29, 2014
This was my research that I worked on in 2012-2013 at the astrophysics lab at UCSC. The work was published in international journals (Harvard Smithsonian, The Astrophysical Journal) before I entered college, I presented my research as a speaker at the 223rd American Astronomical Society Meeting in Washington D.C., and I won several science awards for the work (Intel, Siemens, etc).
This is an HTML cutout from the published article in the Astrophysical Journal which you an access here
. A derived piece of this work was also published in the Harvard/Smithsonian Journal which you can access here
. I wanted to thank Elisa Toloba
, Puragra Guhathakurta
, Aaron Romanowsky
, and Jacob Arnold
for advising me. And of course my longtime friend and lab partner Neel Ramachandran
We present a new spectroscopic technique based in part on targeting the upward fluctuations of the surface brightness for studying the internal stellar kinematics and metallicities of low surface brightness galaxies and streams beyond the Local Group. The distance to these systems makes them unsuitable for targeting individual red giant branch (RGB) stars (tip of RGB at I 24 mag) and their surface brightness is too low (μr 25 mag arcsec−2) for integrated light spectroscopic measurements. This technique overcomes these two problems by targeting individual objects that are brighter than the tip of the RGB. We apply this technique to the star-forming dwarf galaxy NGC 4449 and its stellar stream. We use Keck/DEIMOS data to measure the line-of-sight radial velocity out to ∼ 7 kpc in the East side of the galaxy and ∼ 8 kpc along the stream. We find that the two systems are likely gravitationally bound to each other and have heliocentric radial velocities of 227.3 ± 10.7 km s−1 and 225.8±16.0 km s−1, respectively. Neither the stream nor the near half of the galaxy shows a significant velocity gradient. We estimate the stellar metallicity of the stream based on the equivalent width of its Calcium triplet lines and find [Fe/H] = −1.37 ± 0.41, which is consistent with the metallicity- luminosity relation for Local Group dwarf galaxies. Whether the stream’s progenitor was moderately or severely stripped cannot be constrained with this metallicity uncertainty. We demonstrate that this new technique can be used to measure the kinematics and (possibly) the metallicity of the numerous faint satellites and stellar streams in the halos of nearby (∼ 4 Mpc) galaxies.
The Λ cold dark matter (ΛCDM) cosmological scenario implies that galaxies assemble hierarchically, i.e. the smallest halos contribute to the build up of the most mas- sive ones (e.g., Springel et al. 2006). These smaller halos are severely affected by the larger potential well of the host and are tidally stripped or fully disrupted, appear- ing on the sky as substructures in form of streams and satellites (e.g., Bullock & Johnston 2005; Springel et al. 2006; Cooper et al. 2010). Observationally, the detec- tion of substructures in the halos of massive galaxies agrees with this scenario (e.g., Mart ́ınez-Delgado et al. 2010; Atkinson et al. 2013; Martin et al. 2014). How- ever, there are still discrepancies between observations and simulations that challenge this galaxy formation model. For example, the number of observed satel- lites around massive galaxies is at least one order of magnitude smaller than the predicted number by cos- mological simulations (the so-called “missing satellite problem”, e.g., Klypin et al. 1999; Moore et al. 1999), and the number of dwarf satellites with high circular velocities is significantly smaller than the ΛCDM pre- dictions (the so-called “too big to fail problem”, e.g., Boylan-Kolchin et al. 2011, 2012).
These discrepancies between ΛCDM and observations
are based almost entirely on studies of the satellites and streams found in the Local Group. However, models also predict a large dispersion in the number and prop- erties of the satellites and streams due to varying ac- cretion histories (Johnston et al. 2008) and inhomoge- neous reionization (Busha et al. 2010). Thus, it is im- portant to make a systematic study of these faint struc- tures beyond the Local Group and understand whether these inconsistencies between ΛCDM and observations are a consequence of the Local Group being an outlier or whether some of the physics considered in these sim- ulations need to be adjusted. Some efforts have begun in this direction (Chiboucas et al. 2009, 2013; Sand et al. 2014; Crnojevi ́c et al. 2014; Toloba et al. 2016), but they are generally limited to photometric information. Spec- troscopy of these faint structures and satellites provides dynamical information to model the disruptive events and predict their time scales as well as stellar metallicity information to learn about their star formation histories. Spectroscopy can be carried out using globular clusters and planetary nebulae as bright dynamical tracers (e.g., Foster et al. 2014). However, this approach is feasible only for the brightest dwarfs and streams because the faintest ones have very low numbers of these tracers.
Stellar radial velocity gradients and detailed stellar populations of galaxies have been analyzed using two main techniques: resolved stars and integrated light. The choice between these techniques is based on the distance of the target galaxy and its surface brightness. Low sur- face brightness galaxies within the Local Group, like the satellites of the Milky Way and M31, are studied using resolved stars because their proximity and the brightness of their individual stars allow them to be targeted spec- troscopically with the current telescopes and reasonable integration times (e.g., Geha et al. 2006; Simon & Geha 2007; Kirby et al. 2008a,b; Geha et al. 2010; Kirby et al. 2011; Tollerud et al. 2012; Collins et al. 2013; Ho et al. 2015; Simon et al. 2015). Individual stars in galaxies out- side the Local Group are too faint (I 24 mag) to be tar- geted spectroscopically with current telescopes. In those cases, the technique of integrated light spectroscopy is used. However, this can only be applied to relatively bright objects (μr < 24 mag arcsec−2), which limits the study of spatially resolved galaxy internal properties to nearby objects. Some examples of integrated light spectroscopic studies of spatially resolved dwarf galaxies are reported in Geha et al. (2002, 2003); de Rijcke et al. (2005); Koleva et al. (2009); Toloba et al. (2009, 2011, 2014a,b, 2015); Ry ́s et al. (2013). Here we describe a new spectroscopic technique that allows for kinematic and metallicity studies of faint satellites and streams be- yond the Local Group and up to a distance of ∼ 4 Mpc.
2. New SBF Spectroscopic Technique
We have developed a new technique to spectroscopi- cally target low surface brightness satellites and streams to obtain their internal stellar kinematics and (possi- bly) their metallicity. This technique combines two tra- ditional spectroscopic methods: integrated light spec- troscopy and multi-object spectroscopy of individual stars. The main idea is as follows: instead of placing a long slit across a low surface brightness target (e.g., dwarf galaxy or stellar stream), we break the long slit into multiple short segments (slitlets). These slitlets are then moved along the direction perpendicular to the length of the original long slit so they target the most prominent peaks of the non-uniform surface brightness distribution of the target galaxy or stream.
2.1. Detailed description of the technique
This spectroscopic technique is based on the same phe- nomenon that is the basis of the surface brightness fluc- tuation (SBF) technique used to measure distances. Sur- face brightness fluctuations are caused by Poisson fluc- tuations in the number of stars blended within a seeing disk from one location to another. These fluctuations are the largest for the most luminous stars since their numbers are the smallest and yet their fractional contri- bution to the overall surface brightness is relatively high. The fractional amplitude of SBFs depends on distance, surface brightness, and stellar luminosity function in the following ways: (1) the further away a galaxy, the larger the number of stars in any given seeing disk and the smaller the Poission fluctuations or SBF; (2) the higher the surface brightness of the galaxy, the larger the num- ber of stars in any seeing disk and the smaller the SBF (however, for high surface brightness objects the tradi- tional long slit spectroscopy works better); and (3) the more steeply the star counts rise towards fainter magni- tudes at the bright end of the luminosity function, the larger the SBF. The metallicity distribution function, the age distribution of the galaxy and the choice of the pho- tometric filter can affect how steep the bright end of the luminosity function is.
We combine this new spectroscopy technique with the traditional spectral stacking resulting in two flavors of coaddition. First, since our technique targets upward SBFs, it essentially takes advantage of the natural blends of stars. Stars that are nearly co-spatial in projection are detected as bright blended sources in photometric catalogs derived from images. Of course, the apparent brightness of a blended source is higher than the bright- nesses of the constituent individual stars. Second, we do the usual spectral coaddition of faint sources in order to boost the S/N ratio of the resulting spectrum. Next we describe the spectroscopic target selection and analyze these two different kinds of coaddition in the context of the application of our technique to the recently discov- ered stellar stream near the star-forming dwarf galaxy NGC 4449.
2.2. Photometric selection of spectroscopic candidates
Our new technique is based on targeting objects identi- fied as possible members of the galactic structure under study based on their color-magnitude diagram (CMD) and position in the sky. The objects targeted are those close to the tip of the RGB (TRGB) or brighter. The selection of fainter objects will depend on the telescope and instrument used and the exposure time. These se- lected candidates are blends of RGB stars, and possibly some asymptotic giant branch (AGB) stars if there is an intermediate-age population in the galactic structure un- der study (see Mart ́ınez-Delgado & Aparicio 1997, for a study of the blending effects on synthetic CMDs).
We apply this technique to the stellar stream of NGC 4449 (Mart ́ınez-Delgadoetal. 2012; Richetal. 2012). Due to the distance to NGC 4449 (3.82 Mpc; Annibali et al. 2008), the apparent magnitudes of indi- vidual RGB stars are too faint to target them spectro- scopically and its extremely low surface brightness also prevents us from obtaining integrated light spectroscopy (the TRGB of NGC 4449’s stream is found at I = 24.06 and its surface brightness is μg = 26.75 mag arcsec−2; Mart ́ınez-Delgado et al. 2012). However, this stream, and its companion dwarf galaxy NGC 4449, are resolved into individual and blends of stars in deep photometric images and thus it is a perfect target for testing this new technique.
We use our Subaru/Suprime-Cam r − i CMD, which Mart ́ınez-Delgado et al. (2012) used to confirm the dis- tance of the stream, to select candidates in the stel- lar stream. Our primary targets are objects above the TRGB (i ≤ 24.2) and in the color range −0.1 < r − i < 0.5. A few additional objects with bluer colors and fainter magnitudes are added to fill the DEIMOS slit- mask as ’filler’ targets, in the sense that they use parts of the slitmask area that would otherwise go unused. Fig- ure 1 shows the positions of selected candidates in the CMD.
2.3. Observations and data reduction
We used the DEIMOS spectrograph (Faber et al. 2003) located at the Keck II 10 m telescope in the Mauna Kea Observatory (Hawaii). The stream has an elon- gated morphology that is ∼ 8 kpc long by ∼ 1.5 kpc wide (∼ 6.4′ × 1.4′; Mart ́ınez-Delgado et al. 2012). This size is smaller than the Keck/DEIMOS field of view (16.7′ × 5.0′). To efficiently fill the slitmask, we added some targets in NGC 4449 itself. These targets were selected in the same way as the targets in the stream. In addition, we selected three globular clusters (GCs) from the catalog by Strader et al. (2012) which is based on HST/ACS photometry. These GCs are brighter than 21.8 mag in the HST/ACS F555W filter. We did not tar- get more GCs because we gave higher priority to objects selected in the same way as in the stellar stream. See Figure 2 for an overlay of the designed Keck/DEIMOS mask footprint over NGC 4449 and its stream.
The observations were carried out using the 1200 l/mm grating centered at 7800 ̊A with slit widths of 1.0′′ and the OG550 filter to block shorter wavelength light. All the slits were aligned with the mask position angle (PA= −12◦). This instrumental configuration provides a wavelength coverage of ∼ 6500 − 9000 ̊A with a spec- tral pixel scale of 0.33 ̊A/pixel, and a spectral resolution of 1.4 ̊A (FWHM) or R ∼ 6000. The observations took place on April 4th, 9th, and 10th 2013. The average seeing conditions were 0.6′′, 0.8′′, and 0.8′′ in FWHM, respectively. The total exposure time for this slitmask was 8400 s split in four individual exposures of 1200 s and 2 of 1800 s.
The raw two-dimensional spectra were reduced and ex-tracted into one-dimensional spectra using the DEIMOS spec2d pipeline designed by the DEEP Galaxy Redshift Survey team (Cooper et al. 2012; Newman et al. 2013) and modified by Simon & Geha (2007) to optimize the reduction of resolved targets. The main steps in the re- duction process consisted of flat-field and fringe correc- tions, wavelength calibration, sky subtraction, and cos- mic ray cleaning.
The reduced one-dimensional spectrum was obtained by identifying the target in the reduced two-dimensional spectrum and extracting a small window centered on it. The target was identified by finding the peak of the spatial intensity profile obtained by collapsing the two- dimensional spectrum in the wavelength direction. A Gaussian function was fitted to the target and its width was used as the extraction window. The one-dimensional spectrum was obtained by extracting the spectral rows within this window, weighting by the Gaussian function.
2.4. Tests to identify the nature of our targets
The reliability of the results obtained from this tech- nique depends on the nature of the target objects. Whether our targets mainly consist of several objects with similar luminosity or one luminous and a few fainter objects will affect the accuracy of the radial velocities measured and could dramatically influence the stellar metallicity estimated. If our blends contain objects that are not part of the target galaxy or stream, such as Milky Way (MW) halo and disk stars and background galax- ies, the resulting velocities will be affected, too. In the case of MW stars, their similar radial velocity to that of NGC 4449 can broaden the absorption lines and bias the radial velocities and metallicities. In the case of back- ground galaxies only those with absorption lines will af- fect the resulting measurements. Those with clearly iden- tifiable emission lines within the Keck/DEIMOS spectral window are removed from the sample. The absorption lines of these galaxies will appear at a wavelength that is very different from the absorption lines of NGC 4449 and the stellar stream and thus the radial velocities will not be affected. However, these galaxies could contribute to the continuum, diluting the signal, which results in an underestimation of the metallicity.
Here we perform some tests to analyze the blending effects in our data and to estimate how many MW stars and absorption line background galaxies are expected in our Keck/DEIMOS slitmask.
2.4.1. Stellar blends
The number of blended stars, as described in Section 2.1, depends on the line-of-sight distance, the surface brightness, and the stellar luminosity function. Apart from line-of-sight distance, the other two properties are very different for NGC 4449 versus the stellar stream. While NGC 4449 is a starburst galaxy with young and intermediate-age stellar populations (Annibali et al. 2008; Ry ́s et al. 2011), the stream, with significantly lower surface brightness, mainly contains old stars (∼ 10 Gyr) and possibly a handful of intermediate-age AGB stars (Mart ́ınez-Delgado et al. 2012). Thus, the level of blending is expected to be different in NGC 4449 versus the stellar stream. We analyze each of them separately below.
We study the blending effects on our NGC 4449 tar- gets by comparing our Subaru/Suprime-Cam photome- try with the publicly available HST/ACS image of the central regions of NGC 4449 (Ry ́s et al. 2011). Only a small region of our Keck/DEIMOS slitmask overlaps with the HST image of NGC 4449. The HST image cov- ers only the colored portion of the galaxy image shown in Figure 2. In that small region, we spectroscopically targeted a dozen objects and three GCs. In the case of the GCs, the surrounding objects are several magnitudes fainter and they do not contribute to the GC spectra. We examine the blending of the remaining twelve target objects and show an example in Figure 3.
Only in one out of the twelve objects does the stellar blend consist of more than three stars of similar luminos- ity. This slit is the closest to the center of NGC 4449 in our sample. In this case, the crowding of the stars due to the increase of the stellar density in the central regions of NGC 4449 begins to severely affect the blending. For this reason, we avoided targeting objects in the central regions of the galaxy and targeted objects located more than 1 kpc away from the center of NGC 4449. The S/N of this particular target is too low and we cannot use it to estimate a radial velocity measurement. Therefore, this target is part of the spectral coaddition described in Section 2.5. Including or removing this target from our analysis of radial velocities does not change the results.
In the remaining eleven objects with HST data, there is only one blend that consists of two stars of similar luminosity. The remaining blends consist of a bright star and a few fainter stars (see Figure 3). Thus, the bright star dominates the light in that blend and the radial velocity obtained is not strongly affected by the blending.
There is no HST image available for the stellar stream, thus, we study its stellar blending by analyzing the expected number of blends in our selection box above the TRGB (see the locus of selected objects in Figure 1). We quantify the expected number of blends by simulat- ing CMDs for streams such as the one near NGC 4449 we have observed here. We do this by taking the photome- try of the observed objects in the stream that are below the TRGB and randomly distributing them in space fol- lowing an exponential profile that spans the total size of the stream (∼ 1.5 × 7 kpc). We performed 100 such sim- ulations and, for each one, we count the number of RGB stars that appear blended in one seeing disk avoiding du- plications. We adopt a FWHM 0.8′′ as typical seeing for our observations. Then, we calculate the brightness and the color of these seeing disks (the i band magnitude and r − i color). Finally, we count how many of these blends land within our selection box. The median number and standard deviation of these simulations indicate that we should expect 207 ± 12 blends of RGB stars in our selec- tion box.
However, not all the photometric objects are selected to be a spectroscopic target. Using our Subaru/Suprime- Cam catalog of candidate objects in the Keck/DEIMOS field of view (16.3′ × 5′), we calculate that the fraction of selected objects is ∼ 30%. Even though the area of the Keck/DEIMOS slitmask is not filled uniformly and the stellar stream covers a small fraction of the third chip of DEIMOS, counting the chips from North to South in Figure 2, the selection function is the same because the spatial conflicts do not allow to place more slits in that region. Thus, applying this selection function and taking into account that the stream covers only ∼ 60% of the area of the Keck/DEIMOS slitmask, we expect to have ∼ 37 ± 2 blends of RGB stars in our sample of stream objects above the TRGB.
2.4.2. Background and foreground contamination
We study the kind of objects we target in the stream by estimating the number of contaminants: background galaxies and Milky Way stars. We make this estima- tion in three independent ways: (1) using our photo- metric data to estimate the number of objects that land within our selection box; (2) using very deep HST ob- servations of GOODS-South to estimate the number of compact quiescent galaxies expected in an area the size of the Keck/DEIMOS slitmask. The compact star form-ing galaxies with emission lines within the wavelength coverage of our instrumental configuration are visually identified and removed from the sample; and (3) using the Besanc ̧on model to estimate the number of MW stars expected in the line of sight of NGC 4449 in an area the size of the Keck/DEIMOS slitmask.
In the first test, we estimate the number of contam- inants by counting the number of objects that land in the selection region above the TRGB in an area of our Subaru/Suprime-Cam image that is not affected by the stream or NGC 4449. Assuming that the distribution of contaminants is uniform, we expect to have 366 contam- inants within our selection box. Taking into account the selection function discussed in Section 2.4.1, we expect to have ∼ 110 contaminants in our sample and ∼ 66 of them landing in the area of the stellar stream. These contami- nants can be background galaxies, both star forming and quenched, and foreground stars.
In the second test, we use the CANDELS catalog of GOODS-South galaxies by Guo et al. (2013) to estimate the number of quenched background galaxies expected in a Keck/DEIMOS slitmask. We select galaxies with a di- ameter less than or equal to the typical seeing during our ground-based Subaru/Suprime-Cam observations (0.5′′). These are identified as point sources in our catalogs. We select quenched galaxies as those with U − B > 1.0, which roughly represents the red sequence of galaxies (e.g., Bell et al. 2004; Faber et al. 2007). We finally se- lect objects with i band magnitudes and r−i colors in the same range as our objects (see Figure 1). This selection leads to 0.15 galaxies arcmin−2 in GOODS-South field. Assuming that there is no cosmic variance and taking into account our selection function, we expect to have 3.6 compact quenched galaxies in our sample, of which ∼ 2 of them can land in the region covered by the stellar stream.
In the third test, we apply the Besanc ̧on model (Robin et al. 2003) to the line of sight of NGC 4449, which is located at RA = 12h28m11.1s and DEC = +44o5′37′′, or in Galactic coordinates Longitude = 136.85318o and Latitude = 72.40073o. We use the i band magnitude range and r − i color range covered by our targets (see Figure 1). The expectation is to have 14 MW stars in the full footprint of Keck/DEIMOS. Ap- plying the same selection function and area coverage to the MW stars, we obtain ∼ 4 expected MW contami- nants. This means that we expect to find ∼ 2 MW stars in the area where the stellar stream lands. In our radial velocity analysis we find 1 MW star candidate based on its negative radial velocity. The measured radial velocity of NGC 4449 is 207 km s−1 (Schneider et al. 1992) and the expected velocity dispersion is ∼ 20 km s−1 based on the Large Magellanic Cloud and other dwarf galaxies of similar luminosity (McConnachie 2012). This suggests that an object with a negative velocity is very likely to be a MW star.
2.4.3. Conclusions from the tests
NGC 4449 has a much shallower luminosity function than the stellar stream because of its very prominent young population of stars (Ry ́s et al. 2011). The young stars in NGC 4449 are brighter than the old stars and stand out in any seeing disk. This makes the blending effects for NGC 4449 negligible. The objects targeted above the TRGB for this galaxy will be dominated by intermediate-age AGB stars.
The stellar stream, on the other hand, has a much steeper luminosity function because it is mainly com- posed of old stars (Mart ́ınez-Delgado et al. 2012), which means that the bulk of the stars have very similar lu- minosity. This makes the blending effects in the stellar stream very important. We quantify this blending by counting the number of observed objects in our selec- tion box above the TRGB, which is 651 (based on the CMD shown in Figure 1). Out of these 651 objects, 366 are expected to be background contaminants, 4 are ex- pected to be foreground contaminants, and 207 are ex- pected to be blended RGB stars. Thus, the remaining ∼ 74 targets could be intermediate-age AGB stars. Tak- ing into account the selection function and the area of the Keck/DEIMOS slitmask covered by the stream, we expect ∼ 13 intermediate-age AGB stars in our sam- ple of 51 stellar stream objects above the TRGB. The majority of the contaminants are expected to be com- pact background star-forming galaxies, the number of expected quenched galaxies and MW stars in this area is negligible. These star forming galaxies are easily iden- tified by their emission lines and are removed from our sample.
In summary, our targets in NGC 4449 are mainly in- dividual intermediate-age AGB stars and in the stellar stream our targets are dominated by blends of RGB stars but they can also contain some individual intermediate- age AGB stars.
2.5. Spectral coaddition
Because of the faintness of the objects targeted, not all the spectra have identifiable absorption lines and the radial velocity measurements are not always reliable. To improve the quality of the data we coadd several spectra in groups. Our grouping scheme is based on the target position on the sky and we require our groups to have a similar number of spectra. As a result, each group has a minimum of fifteen objects and the resulting S/N is at least ∼ 2 ̊A−1 in the Calcium triplet region. In the case of the stellar stream, we group objects that are close together along the stream. In the case of NGC 4449’s main body, we group objects that are at a similar pro- jected distance with respect to the center of the galaxy (see Figure 2).
The spectral coaddition process is based on the follow- ing steps: (1) rebinning the spectra and their associated uncertainties obtained in the reduction process onto a common wavelength range (6300 − 9000 ̊A) – because the data come from a mask of slits, the spectral range of each slit depends on its position within the mask; (2) renormalizing the fluxes and their associated uncertain- ties; and (3) adding, pixel by pixel, the fluxes of the nor- malized rebinned spectra weighted by the scaled version of the rebinned inverse variance (the inverse variance is scaled to match the normalization of the flux). This ad- dition is performed after doing a sigma clipping where those pixels that deviate more than 3σ from the median are rejected and not added. The distance with respect to NGC 4449 of each coadded spectrum is the average distance of all the objects included in that particular coaddition. This coaddition technique works best when the internal velocity dispersion of the target galaxy or stream is lower than the velocity uncertainties (typically 20 − 30 km s−1; see Table 1 and Section 3). We assume this to be the case here for NGC 4449 and its stellar stream. The expected velocity dispersion for NGC 4449 is∼20kms−1 andforthestreamis 20kms−1 given their very low luminosity (e.g., McConnachie 2012).
Figure 4 shows an example of a coadded spectrum and some of the individual spectra that go into that coaddi- tion. The individual spectra before being coadded are very noisy and have very few identifiable features. How- ever, when they are coadded together, the features that were very subtle or not identifiable in some of the spectra appear more clearly and can be used to measure a radial velocity.
3. Summary (Paper has further conclusions)
We present a new spectroscopic technique to measure the stellar kinematics and metallicities of diffuse objects that are beyond the Local Group. This technique con- sists of targeting objects that are above the tip of the RGB. These objects are a mixture of some AGB stars, if an intermediate-age population is present in the target galaxy or stream, and blends of RGB stars that appear brighter because of their spatial superposition in the sky. Given the brightness of these objects (I 23.5 mag) we can measure their line-of-sight radial velocities and, coadding their individual spectra, we can also estimate the metallicity of the target galaxy or stream. The metal- licity measurements are limited to old stellar populations where intermediate-age AGB stars are not present or their number is negligible. This is because of the lack of a calibration to transform CaT EWs into metallicities. This limitation will no longer exist when these calibra- tions are established.
We apply this technique to the nearby star form- ing dwarf galaxy NGC 4449 and its stellar stream lo- cated at 3.8 Mpc (Annibali et al. 2008). We measure the line-of-sight radial velocity along the galaxy and the stream and find flat radial velocity gradients for both systems, within our velocity uncertainties and the fact that for the galaxy we only sample its East side and it is not along the semimajor axis. We also find that the median heliocentric velocity of both systems are very similar to each other, V = 227.3 ± 10.7 km s−1 forNGC4449andV=225.8±16.0kms−1forthe stream, which suggests that the two systems are grav- itationally bound. We estimate the global metallicity of the stellar stream because it is mainly composed old stars, while our targets for NGC 4449 seem to be mainly intermediate-age AGB stars. Our measured metallicity for the stream is [Fe/H] = −1.37 ± 0.41, which is con- sistent with the metallicity-luminosity relation of Local Group dwarf galaxies. However, the uncertainty of the metallicity is too large to constrain whether the progeni- tor of the stream was a dwarf galaxy with properties very similar to those dwarfs found in the halo of the Milky Way or M31, or, on the contrary, it was a larger galaxy that was severely stripped by NGC 4449, as it happened for the Sagittarius stream which has the same luminosity and metallicity as the stream.
We demonstrate that this new SBF spectroscopic tech- nique is a powerful method for studying the kinemat- ics and metallicities of the wealth of dwarf faint satel- lites and streams that are currently being discovered in the local universe. We will apply this same technique to the newly discovered dwarf satellites of M81 located at ∼ 3.8 Mpc (Chiboucas et al. 2009, 2013) and to the dwarf galaxies and streams that we are finding within the PISCeS survey (Panoramic Imaging Survey of Centau- rus and Sculptor; Sand et al. 2014; Crnojevi ́c et al. 2014, 2015; Toloba et al. 2016). These data provide ingredients for detail modeling of the collision between the progeni- tor galaxy and the host galaxy, and also provide obser- vational constraints on the host’s potential well and on the properties and orbit of the progenitor of the streams and satellites.