Quantifying the Faint Structure of Galaxies: The Late-type Spiral NGC 2403
Mike K. Barker, Annette M. N. Ferguson, Mike J. Irwin, Nobuo Arimoto, Pascale Jablonka
aa r X i v : . [ a s t r o - ph . C O ] S e p Mon. Not. R. Astron. Soc. , 1–20 (2010) Printed 25 February 2018 (MN L A TEX style file v2.2)
Quantifying the Faint Structure of Galaxies: The Late-typeSpiral NGC 2403 ⋆ † Michael K. Barker , Annette M. N. Ferguson ‡ , M. J. Irwin , N. Arimoto , ,P. Jablonka , SUPA, Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh, UK, EH9 3HJ Institute of Astronomy, Cambridge University, Cambridge, UK, CB3 0HA National Astronomical Observatory of Japan, Mitaka, Tokyo 181-8588, Japan Department of Astronomical Science, Graduate University for Advanced Studies, Mitaka, Tokyo 181-8588, Japan Laboratoire d’Astrophysique, Ecole Polytechnique F´ed´erale de Lausanne (EPFL), Observatoire, CH-1290 Sauverny, Switzerland GEPI, Observatoire de Paris, CNRS UMR 8111, Universit´e Paris Diderot, F-92125, Meudon, Cedex, France
Accepted —- 12 September 2011. Received —-; in original form —-
ABSTRACT
Ground-based surveys have mapped the stellar outskirts of Local Group disc galax-ies in unprecedented detail, but extending this work to other galaxies is necessary inorder to overcome stochastic variations in evolutionary history and provide more strin-gent constraints on cosmological galaxy formation models. As part of our continuingprogram of ultra-deep imagery of galaxies beyond the Local Group, we present awide-field analysis of the isolated late-type spiral NGC 2403 using data obtained withSuprime-Cam on the Subaru telescope. The surveyed area reaches a maximum pro-jected radius of 30 kpc or deprojected radius of R dp ∼
60 kpc. The colour-magnitudediagram reaches 1.5 mag below the tip of the metal-poor red giant branch (RGB) ata completeness rate >
50% for R dp &
12 kpc. Using the combination of diffuse lightphotometry and resolved star counts, we are able to trace the radial surface brightness(SB) profile over a much larger range of radii and surface brightness than is possiblewith either technique alone. The exponential disc as traced by RGB stars dominatesthe SB profile out to & R dp ∼
18 kpc, and reaches a V -bandSB of µ V ∼
29 mag arcsec − . Beyond this radius, we find evidence for an extendedstructural component with a significantly flatter SB profile than the inner disc andwhich we trace to R dp ∼
40 kpc and µ V ∼
32 mag arcsec − . This component can befit with a power-law index of γ ∼
3, has an axial ratio consistent with that of the innerdisc and has a V-band luminosity integrated over all radii of 1–7% that of the wholegalaxy. At R dp ∼ −
30 kpc, we estimate a peak metallicity [M/H] = − . ± . α − element enhancement. Although the extantdata are unable to discriminate between stellar halo or thick disc interpretations ofthis component, our results support the notion that faint, extended stellar structuresare a common feature of all disc galaxies, even isolated, low-mass systems. Key words: galaxies: individual: NGC2403 – galaxies: haloes – galaxies: discs –galaxies: structure – galaxies: stellar content ⋆ Based on data collected at the Subaru telescope, which is op-erated by the National Astronomical Observatory of Japan. † Based on observations made with the NASA/ESA HubbleSpace Telescope, obtained from the Data Archive at the SpaceTelescope Science Institute, which is operated by the Associationof Universities for Research in Astronomy, Inc., under NASA con-
The stellar outskirts of galaxies are important testinggrounds for models of galaxy formation and evolution. This tract NAS 5-26555. These observations are associated with pro-gram GO10523. ‡ [email protected] (cid:13) Barker et al. is because the dynamical and star formation timescales thereare relatively long making it easier to identify accreted ma-terial and to study relatively unprocessed gas. N-body andhydrodynamical simulations of galaxy formation within acosmological context predict that the merging and accretionthat is more common at high redshift can leave an imprinton galaxy outskirts that is visible to the present day in theform of thick discs, stellar haloes and discrete substructures(e.g. Brook et al. 2004; Bullock & Johnston 2005; Cooperet al. 2010). Within this scenario, the properties of thesestructures may correlate with host galaxy properties, likepresent-day total mass, but they are also expected to ex-hibit significant variations at the same mass scale due tostochastic variations in the merging/accretion history andthe detailed nature of the individual progenitor systems (e.g.Cooper et al. 2010; Purcell et al. 2007). Thick discs andhaloes may also arise from other processes besides merging,such as radial migration, misaligned gas accretion, and in-situ star formation (Sch¨onrich & Binney 2009; Roˇskar et al.2010; Loebman et al. 2010). Therefore, it is crucial to studyas many galaxies and galaxy types as possible to overcomestochastic variations and discern underlying trends that mayhelp to isolate the dominant formation mechanisms.These outer stellar structures are very faint, typicallyseveral magnitudes below the sky level. Detecting their dif-fuse light requires very careful treatment of sky subtrac-tion, flat-fielding errors, detector response, scattered light,and PSF wings (e.g. Morrison et al. 1997; de Jong 2008).Nevertheless, a growing body of diffuse light analyses sup-ports the idea that such structures are common around discgalaxies (e.g. Zibetti et al. 2004; Zibetti & Ferguson 2004;Jablonka et al. 2010; Burstein 1979; Tsikoudi 1979; Shaw& Gilmore 1990; de Grijs & van der Kruit 1996; Morrisonet al. 1997; Neeser et al. 2002; Dalcanton & Bernstein 2002;Malin & Hadley 1997; Shang et al. 1998; Mart´ınez-Delgadoet al. 2009). Most of these studies have imaged in a singleband hence there is no information on the the nature of theextended stellar populations. Even in cases where multiplepassbands have been obtained, the age-metallicity degener-acy present in optical broadband colours enables only verycrude constraints.An alternative approach to studying galaxy outskirts iswith resolved stars, a technique which can typically reachfar fainter surface brightness (SB) levels than diffuse light.The most interesting cosmological constraints come from theold stars in these systems, those on the red giant branch(RGB). With ground-based telescopes, resolving RGB starsin external galaxies was initially limited to systems withinthe Local Group (e.g. Ferguson et al. 2002, 2007; Ibata et al.2007; McConnachie 2009). These studies found a wealth ofvery faint stellar structures around the MW-analog M31 andcomparatively little around the late-type spiral M33. Thesestructures exhibited large-scale inhomogeneities in distribu-tion and composition, highlighting the importance of arealcoverage when looking at galaxy outskirts. However, theseare just two systems; more rigorous tests of cosmologicalgalaxy formation models require similar data for many moregalaxies beyond the Local Group.With this motivation in mind, we are conducting aprogram to explore the low surface brightness outer re-gions of all large galaxies within 5 Mpc using wide-field im-agers on 8-m class telescopes. In our first paper, we used Subaru/Suprime-Cam to identify an extended structure ofRGB stars around the MW-analog, M81, stretching out to adeprojected radius R dp = 44 kpc (Barker et al. 2009). Thisstructure had a flatter radial and azimuthal surface densityprofile than the main disc suggesting it was a halo or thickdisc, but its properties did not exactly match either of thesecomponents in the MW. Furthermore, as M81 is part of aninteracting group of galaxies, we could not exclude the hy-pothesis that the extended component was the result of arecent tidal encounter.In parallel with our efforts, other groups have pur-sued similar studies but typically with smaller field-of-view(FOV) imagers (e.g. Vlaji´c et al. 2009, 2011; de Jong et al.2008; Rejkuba et al. 2009). Such studies risk being affectedby the presence of localized substructures which fall withinthe FOV and also suffer from significant uncertainty in thebackground/contaminant subtraction which is a crucial as-pect of quantifying low surface brightness emission. Evenfor systems beyond the Local Group, the FOVs of HSTand GMOS are too small to reveal a global picture of theirouter structures and wide-field imagers like Suprime-Camare clearly needed (e.g. Mouhcine et al. 2010; Bailin et al.2011; Tanaka et al. 2011). Perhaps it is not so surprisingthen, that, until recently, no detailed global analysis of RGBstars had been conducted for a spiral galaxy outside theLocal Group.As part of our program, we present here observationsof the late-type Sc galaxy, NGC 2403. With a total mass ∼ M ⊙ and a circular velocity of ∼
135 km s − (Frater-nali et al. 2002), NGC 2403 is similar in many respects toM33 and NGC 300, and is, therefore, a good system withwhich to increase the observed baseline in galaxy mass. Inthis work, we adopt the HST Key Project Cepheid-baseddistance modulus of ( m − M ) = 27 . ± .
24, giving ita distance of 3.13 Mpc (Freedman et al. 2001). This dis-tance compares favorably to HST RGB tip-based distancesof 3.09 – 3.20 Mpc (Dalcanton et al. 2009). At this distance,1 . ′ ≈ . ∼ ◦ from the plane ofthe MW with a Galactic longitude of ∼ ◦ . It has a B-bandisophotal radius of 10 . ′ , or 9.8 kpc, and a photographic V-band disc scale-length h = 1 . m V = 8 .
04 (de Vaucouleurset al. 1991), which translates to an absolute magnitude of M V = − .
44. With an inclination of 63 ◦ (Fraternali et al.2002), inclination-dependent extinction effects are not likelyto be significant ( ∼ . ∼ × M ⊙ and a dynamical mass ∼ M ⊙ inside a radius of 23 kpc. These observationsrevealed no signs of interaction in the immediate vicinity.However, they did reveal an asymmetric warp in the outerdisc and a thick, clumpy layer of HI that rotates more slowlythan, and contains roughly 10% the mass of, the cold HI disc.NGC 2403 is the brightest member of a loose galaxygroup that shows no clear signs of interactions. Chynowethet al. (2009) detected no HI clouds in this group with massesdown to a limit of 2 . × M ⊙ . The nearest systems to NGC2403 are the small galaxies, DDO 44 and NGC 2366, eachlying ∼
80 kpc and ∼
200 kpc away in projection, respec-tively. Five other small galaxies are located ∼
350 kpc away c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 1.
Digitized Sky Survey image showing the sizes andlocations of the two fields observed with Suprime-Cam (F1 andF2). The ellipse marks the R radius of 10 . ′ or 9 . § in projection (Karachentsev et al. 2002). On larger scales,NGC 2403 belongs to a filament of ∼
60 known galaxiesspanning roughly 25 ◦ on the sky. M81 is located near thecenter of the filament ∼
800 kpc from the spirals NGC 2403and NGC 4236, which lie at opposite ends. NGC 2403 is atleast 4 times farther away from the nearest large disc galaxy(M81) than M33 is from M31 and hence it can be considereda far more isolated system.This paper is organized as follows. In §
2, we describethe observations and data reduction procedures. In §
3, wepresent the colour-magnitude diagram (CMD) and, in § § § §
5, the SB profile is derived and we fit the SB profilewith several models and, in §
6, we discuss the implications.Finally, we summarize the results in § The observations were obtained with the Suprime-Cam in-strument (Miyazaki et al. 2002) on the 8-m Subaru telescopeon the night of January 8, 2005 (S04B, PI=N. Arimoto).This instrument consists of 10 CCDs of 2048x4096 pixelsarranged in a 2x5 pattern, with a pixel scale of 0.2 arcsecand a total field of view of approximately 34 ′ x27 ′ (includinglong edge inter-chip gaps of 16 – 17 arcsec and short edgegaps of 5 – 6 arcsec).NGC 2403 was covered using two field centers,one to the north of its nucleus at ( α J , δ J ) =(7 h m s . , +65 ◦ ′ ′′ ) and another to the south at ( α J , δ J ) = (7 h m s . , +65 ◦ ′ ′′ ). The largerectangular boxes in Fig. 1 outline the locations of thesefields, which we refer to as F1 and F2, respectively. For eachfield, we obtained a set of 8 images in the Johnson V fil-ter with individual exposure times of 450s, and 12 imagesin the Sloan i ′ filter with exposure times of 205s. All obser-vations were recorded under slightly non-photometric con-ditions through patchy thin cirrus. The images of F1 weretaken in an average seeing of ∼ . ′′ in both filters while theF2 images had an average seeing of ∼ . ′′ and ∼ . ′′ inthe V -band and i ′ -band, respectively.To fill in the chip gaps and facilitate the removal ofcosmic rays and bad pixels, individual images were ditheredby ∼
25 arcsec, resulting in a mosaic for each field covering ≈ ′ x28 ′ , or ≈ ≈ .
54 deg ,or ≈ , in the central 39x48 kpc around NGC 2403.Flat-field and interchip gain variations were removed withmaster flats obtained by combining 12 and 11 twilight skyflats in the V and i ′ filters, respectively. After flat-fielding,remaining large scale variations in dark sky level, measureddirectly from stacked dark sky images at several differentpositions obtained during this run, were less than 1% of sky.An i ′ -band fringe frame acquired from an earlier Suprime-Cam imaging run was used to help assess the degree of darksky fringing present, but this was found to be negligiblein our data, so this extra image processing step was notrequired.The image processing procedure was very similar to thatfollowed by Barker et al. (2009). After converting the rawdata to multi-extension FITS format, all images and cali-bration frames were run through a variant of the data re-duction pipeline developed for the Isaac Newton Telescope(INT) Wide Field Survey (WFS) . Here we present a briefoverview of the main steps of the pipeline which is describedin more detail in Irwin (1985, 1997); Irwin & Lewis (2001)and Irwin et al. (2004).Prior to deep stacking, catalogues were generated foreach individual processed science image to both refine theastrometric calibration and asses the data quality. Forastrometric calibration, a Zenithal polynomial projection(Greisen & Calabretta 2002) provided a good prescriptionfor the World Coordinate System (WCS) and included allthe significant telescope radial field distortions. We used thisin conjunction with a 6-parameter linear plate model perdetector to define the required astrometric transformations.The 2MASS point source catalogue (Cutri 2003) was usedfor the astrometric reference system.The individual image qualities were then assessed usingthe average seeing and ellipticity of stellar images, as wellas sky level and sky noise determined from the object cata-loguing stage. Images were stacked at the detector level us-ing the updated WCS information to accurately align themto a reference image. The background level in the overlaparea between each image in the stack and the reference wasadjusted additively to compensate for sky variations dur-ing the exposure sequence and the final stack included see-ing weighting, confidence (i.e. variance) map weighting, andclipping of cosmic rays. ∼ wfcsur/c (cid:13) , 1–20 Barker et al.
Figure 2.
Colour composite image of NGC 2403 created from the stacked mosaics and cropped to ≈
24 kpc on a side. The intensityscaling is logarithmic and the colour mapping is similar to that of Lupton et al. (2004) with V for the blue channel, i ′ for red, and theaverage for green. Luminous main sequence and evolved giant stars can be seen to resolve in the outer parts of the disc. North is up andeast is to the left. Next, we generated detector-level catalogues from thestacked images and updated the WCS astrometry in theFITS image extensions prior to mosaicing all detectors to-gether. Residual astrometric errors over the whole stackedarray were typically < . ′′ , greatly simplifying this process.Slight offsets in underlying sky level between the stackeddetector images caused small (typically ∼ . − .
2% ofsky), but still visible, discontinuities in the final mosaics.These offsets were due to small colour equation differencesin the detectors and the relatively blue colour of the twi-light sky compared to dark sky and unresolved diffuse lightfrom NGC 2403. We corrected these offsets iteratively byvisual inspection of a 4x4 blocked version of the mosaics us-ing a pre-assigned keyword in each relevant detector FITSextension designed for this purpose.Fig. 2 shows a colour composite image made from the V and i ′ stacked mosaics of both fields and cropped to ≈
24 kpcon a side. In this image, north is up and east is to the left.The intensity scaling is logarithmic and the colour mappingis similar to that of Lupton et al. (2004) with V for the bluechannel, i ′ for red, and the average for green.The data reduction pipeline also produced an aperturephotometry catalogue using a “soft-edged” aperture withradius close to the full-width-half-maximum (FWHM) (e.g.,Irwin 1997; Naylor 1998). A series of apertures ranging from1/2 to 4 times the FWHM were additionally used to computestellar aperture corrections and to correct for PSF variationacross the field-of-view.The photometric calibration was based on a compar-ison with 0.29 deg of INT V, i ′ -band photometry centredon NGC 2403 taken in photometric conditions during April2009. The INT photometry were converted to the Johnson- c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 3.
CMD of all point sources in F1 and F2. The contoursindicate [140, 200, 260, 320, 380, 440] stars decimag − . Cousins
V, I system using the transformation equationsgiven on the WFS website. All the INT data were calibratedon a nightly basis against multiple observations of Landoltstandard stars. Both Subaru fields for NGC 2403 were com-pared independently with the INT data and were found tohave the same zero-points to within their errors (1 − ∼ .
02 mag.
Because the data reduction pipeline did not provide a di-rect estimate of the completeness rate, we elected to addi-tionally perform PSF-fitting photometry on the stacked mo-saics. This was accomplished using the standalone versionsof the DAOPHOT/ALLSTAR/ALLFRAME suite of pro-grams (Stetson 1987; Stetson & Harris 1988; Stetson 1994).The PSF for each field in each filter was built by startingwith an initial list of roughly 1000 bright, fairly isolatedstars and iteratively subtracting neighbors, rejecting starswith large residuals, and increasing the spatial complexityof the PSF as a function of position on the mosaics. In theend, this left several hundred stars to build a PSF that var-ied quadratically with position on each stacked mosaic.We derived a coordinate transformation between thestacked V and i ′ mosaics of each field using DAOMASTER(Stetson 1993). The stacked mosaics were then coadded withMONTAGE2 (Stetson 1994). Objects meeting a 3 σ detec-tion threshold on the coadded image were measured with Figure 4.
Completeness rate as a function of input magnitude asderived from artificial star tests. The curves show the complete-ness for two radial ranges, R dp <
18 kpc (dashed) and R dp > − I) colour ranges, 0 – 1 (black), 1 – 2 (blue), and 2 – 3(orange).
ALLSTAR to obtain a first guess at positions and magni-tudes and the resulting list was input into ALLFRAME to-gether with the stacked mosaics in each filter and their coor-dinate transformation. To reduce contamination from non-stellar sources in the ALLFRAME photometric catalogue,we excluded objects with abnormally low or high sharp val-ues as measured in the i ′ -band. For the overlap region be-tween F1 and F2 we used only the F1 catalogue since it hadbetter seeing. The PSF magnitudes were standardized to theJohnson-Cousins system using linear transformations to thecalibrated V, I
Subaru aperture photometry described in § − .There are several clear stellar sequences visible, which wediscuss in more detail in § To estimate completeness, we performed artificial star testsin which ∼ c (cid:13) , 1–20 Barker et al.
Figure 5.
The difference between the recovered and input I -band magnitudes of the artificial stars in both fields combined.The lines show the median shifts and the central 68% for all stars(solid) and those with R dp <
18 kpc (dotted) and with R dp >
18 kpc (dashed). There are no significant systematic magnitudeshifts for I . . a total catalogue of ∼ R dp ) . Thus, we show thecompleteness for two radial ranges, R dp <
18 kpc (dashed)and R dp >
18 kpc (solid), and for three different input (V − I)colour ranges, 0 – 1 (black), 1 – 2 (blue), and 2 – 3 (or-ange). Table 1 lists the 50% completeness levels in the I -band for both fields individually and for the total catalogue.In F1, the 50% completeness level occurs at ∼ . − . R dp <
18 kpc and at ∼ . − . R dp >
18 kpc.In F2, the 50% completeness level occurs at ∼ . − . R dp <
18 kpc and at ∼ . − . R dp >
18 kpc.The 50% level is fainter in F1 than it is in F2 because ofthe difference in seeing, but the effect is most noticeable inthe bright optical disc where the crowding is highest. We ac-count for this difference whenever necessary by treating thecompleteness corrections separately for each field. In § Deprojected radii refer to the circular radii within the disc planeand are calculated assuming the 2MASS near-infrared isophotalcenter of ( α J , δ J ) = (7 h m s . , +65 ◦ ′ ′′ ) (Jarrettet al. 2003), an inclination of 63 ◦ , and a position angle of 124 ◦ measured north through east (Fraternali et al. 2002). Figure 6.
The difference between the recovered and input coloursof the artificial stars in both fields combined. The lines show themedian shifts and the central 68% for all stars (solid) and thosewith R dp <
18 kpc (dotted) and with R dp >
18 kpc (dashed).There are no significant systematic colour shifts for I . . Table 1. I -band 50% completeness levels.( V − I ) = 0 – 1 1 – 2 2 – 3 R dp <
18 kpcField F1 24.2 23.9 23.5Field F2 23.1 23.2 22.9Total catalogue 23.5 23.5 23.4 R dp >
18 kpcField F1 25.9 25.5 24.9Field F2 25.3 25.0 24.4Total catalogue 25.6 25.3 24.7
In Fig. 5, we show the I -band photometric shifts of theartificial stars, ∆ I , for the total catalogue. This shift is de-fined as the recovered magnitude minus the input magni-tude. The lines show the median shift and the central 68%for stars at all radii (solid) and those with R dp <
18 kpc(dotted) and with R dp >
18 kpc (dashed). For all radii, themedian of | ∆ I | is ≈ . I = 24 . ≈ . I = 25 . V − I ). For all radii, the median of | ∆( V − I ) | is ≈ . I = 24 . ≈ . I = 25 .
7. These figures showthat there are no significant systematic colour or magnitudeshifts for I . . c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 7.
Point source CMD with theoretical isochrones fromMarigo et al. (2008) overlaid. The young isochrones have ages of10.0, 17.8, 31.6, 56.2, 100, and 178 Myr and a metallicity [M/H]= − .
4. The three old isochrones have a common age of 10 Gyrand [M/H] = –1.3, –0.7, and –0.4. The boxes are used to selectstars in different evolutionary stages: main sequence and bluehelium burning (MS+BHeB), red supergiant (RSG), asymptoticgiant branch (AGB), and red giant branch (RGB). The error barson the right-hand side show typical photometric errors derivedfrom artificial star tests.
In Fig. 7, we show the de-reddened point source CMD withisochrones from Marigo et al. (2008) shifted to the distanceof NGC 2403. On the right-hand side are typical photomet-ric errors (the median of | ∆ I | ) from the artificial star tests.Clearly, there is a range of ages and metallicities present inthese fields. The young isochrones at ( V − I ) ∼ ≈ log( Z/Z ⊙ ) = − .
4. This metallicity should berepresentative of the young populations in the disc, as Gar-nett et al. (1997) measured [O/H] of HII regions to decreasefrom roughly solar at 1 kpc to about 0.4 dex below solar at6 kpc. The three old isochrones at ( V − I ) ∼ − E ( B − V ) value of 0.04 with negligible spatial vari-ation. We adopt the Cardelli et al. (1989) extinction law,for which R V = 3 . A I /A V = 0 . ∼ Figure 8.
CMD of foreground stars based on the Besan¸con modelof the Milky Way (Robin et al. 2003) for a field with the samelocation and total area as F1 and F2. The stars’ magnitudes werescattered according to the artificial star tests, but no completenesscorrections were applied. The number of predicted foregroundstars is ∼
8% of the number of point sources over the observedmagnitude range. The selection boxes avoid the most heavily con-taminated regions. the Schlegel et al. (1998) maps and replaced with medianvalues from the surrounding sky. Thus, the star-by-star cor-rection does not include extinction internal to NGC 2403,but this should not be a serious problem since (i) we mainlyfocus on regions outside the bright optical disc, (ii) we aremostly concerned with the RGB stars, which tend to liefarther away from the high extinction star forming regionsthan young stars (Zaritsky 1999), and (iii) our CMD selec-tion boxes are large compared to the expected amount ofinternal extinction.In what follows, we will focus on several particular CMDregions (outlined in Fig. 7) which isolate stars in differentevolutionary stages in NGC 2403. The blue lines in Fig. 7mark the region occupied by young main sequence and bluehelium burning stars (MS+BHeB) with ages ∼ −
150 Myr.Stars within the cyan polygon are red supergiants (RSGs)with ages in the range ∼ −
180 Myr. The red lines encloseRGB stars, which can have ages ∼ −
10 Gyr. The brightedge of the RGB box is set by the RGB tip of the isochroneswhich has a small metallicity dependence. There could besome contamination of the RGB box by AGB stars and byyoung, red helium burning stars with masses of ∼ − M ⊙ ,particularly if they have [M/H] > − .
4. The magenta linesenclose AGB stars above the RGB tip, which tend to havesomewhat younger ages ( ∼ . − c (cid:13) , 1–20 Barker et al.
Figure 9.
CMD of extended objects. The selection boxes avoidthe most heavily contaminated region. The total number of ex-tended objects is approximately equal to the number of pointsources. ([M/H] . − .
7) and “metal-rich” ([M/H] & − .
7) subre-gions. There will be some overlap in the metallicities probedby these subregions due to photometric errors, but theyare broader than the errors, and so are useful in identify-ing any population gradients. Some of the most metal-poor([M/H] . − .
3) and metal-rich ([M/H] & − .
4) RGB starsmay fall outside the total RGB box, but extending it fur-ther to the blue or red would increase contamination fromMW foreground stars and unresolved background galaxies,and would increase uncertainties due to incompleteness. Wenote that our use of the term, metal-poor, differs somewhatfrom the traditional sense because it includes metallicitiesup to [M/H] = − . I ∼
25, the peak in the colour distribution moves towardbluer colours because of the increasing contamination fromunresolved background galaxies. To mitigate this contami-nation, we do not use any sources fainter than I = 25.Fig. 8 shows the foreground star CMD predicted by theBesan¸con model of the MW (Robin et al. 2003) for the sametotal area and line of sight as our observations. We appliedextinction corrections to the foreground stars in the sameway as for the real data. The stellar colours and magnitudeshave been scattered using a simple exponential function tomimic the increase of photometric error with magnitude seenin the artificial star tests. The Besan¸con model predicts that the number of foreground stars is ∼
8% of all point sourcesover the magnitude limits of the NGC 2403 CMD. Thereare two main features in the foreground star CMD, a nar-row vertical strip at ( V − I ) ∼ . − . I &
20 and in the MW disc at I .
20 whilethe curved band is comprised of late-type main sequencestars in the MW disc. These two features are visible in theNGC 2403 CMD at magnitudes brighter than I ∼
22. Ascan be seen in Fig. 8, the CMD selection boxes sample re-gions that minimize contamination from foreground stars.To check the effectiveness of our morphological classi-fication, Fig. 9 shows the CMD of the objects classified asextended (i.e., with high sharp values indicating a poor fitto the stellar PSF). The CMD selection boxes and contoursare overlaid to facilitate comparison with Fig. 7. The num-ber of extended objects is about the same as the number ofpoint sources. The extended objects are concentrated in abroad diagonal band, the bulk of which lies at bluer coloursthan the RGB and AGB selection boxes. Importantly, theextended objects have a colour-magnitude distribution thatis different from that of the point sources, and most of thestellar sequences in Fig. 7 are not visible. Examination ofthe extended object spatial distribution in the sky revealsthat some of them are misclassified stars located in heav-ily crowded regions in the bright optical disc of NGC 2403.However, these are the most poorly measured objects andthey should not affect our conclusions because we apply com-pleteness corrections to the radial star count profiles and werely on the diffuse light profile inside R dp ∼ In Fig. 10, we plot the two-dimensional spatial distribu-tion of sources in the CMD selection boxes. Going clock-wise from top left, the maps show RGB, AGB, RSG, andMS+BHeB point sources. No correction for contaminantshas been made. The ellipses correspond to R dp = 10 −
60 kpcin steps of 10 kpc (1 kpc ≈ . ′ ). The hole in the nucleus isdue to severe stellar crowding. A few highly saturated starsalso appear as smaller holes. The horizontal white stripesat η = ± . c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 10.
Going clockwise from top left, the tangent plane projection of RGB stars (ages ∼ −
10 Gyr), AGB stars (ages ∼ . − ∼ −
180 Myr), and MS+BHeB stars (ages ∼ −
150 Myr). Ellipses denote deprojected radii of 10 – 60 kpcin steps of 10 kpc (1 kpc ≈ . ′ ). No correction for contaminants has been made to these maps. The hole in the nucleus is due to severestellar crowding. A few highly saturated stars also appear as smaller holes. The horizontal white stripes at η = ± . Fig. 11 shows the raw surface density profiles for pointsources in the CMD selection boxes (i.e., before any com-pleteness correction or background subtraction has been ap-plied). Low confidence pixels near the chip edges and mosaiccorners are excluded from the profiles. The lines are colour-coded so that the total RGB box is red, metal-poor RGBbox is orange, metal-rich RGB box is green, AGB box is ma- genta, MS+BHeB box is blue, and RSG box is cyan. The topprofile is the total of all the boxes. Each point in the profilesis the mean R dp of all stars within a bin. Horizontal errorbars span the full radial range of stars in each bin. Verticalerror bars include Poisson noise, which may underestimatethe true error because it does not include background galaxyclustering. Severe crowding causes the profiles to turn overat R dp ∼ −
10 kpc. Beyond 10 kpc, there are two regimes c (cid:13) , 1–20 Barker et al.
Figure 11.
Raw star count profiles of point sources in the CMDselection boxes. No completeness correction or background sub-traction has been applied. Vertical error bars include Poissonnoise. Horizontal error bars span the full radial range of starsin each bin. Severe crowding causes the profiles to turn over nearthe nucleus at R dp ∼ −
10 kpc. Beyond 10 kpc, there are tworegimes visible, one that extends out to ∼
18 kpc where the pro-files have a steep slope and another beyond 18 kpc where theyare relatively flat. The metal-poor RGB profile decreases out to ∼
40 kpc. visible, one that extends out to ∼
18 kpc where the profileshave a steep slope and another beyond 18 kpc where theyare relatively flat. In particular, the metal-poor RGB profiledecreases slowly out to ∼
40 kpc showing the first indica-tion of an outer structure of metal-poor RGB stars aroundNGC 2403.Next, we correct the raw star counts for completeness byweighting each star by w j = 1 /c j , where c j is the star’s com-pleteness interpolated in colour, magnitude, and R dp usingthe artificial stars in the appropriate field. Fig. 12 shows theweighted mean completeness rate (i.e., Σ w j c j / Σ w j ) of theCMD selection boxes in the total catalogue. Beyond 18 kpc,the completeness is approximately constant and inside 18kpc, the completeness drops with decreasing radius becauseof stellar crowding. This figure shows that the metal-poorRGB has a completeness rate >
50% for R dp &
12 kpc. Sim-ilarly, the total profile is >
50% complete for R dp &
11 kpc.Because of the difference in seeing between the two fields,the total profile is >
50% complete for R dp & § Figure 12.
Completeness rate as a function of R dp for the CMDselection boxes. Beyond 18 kpc, the completeness is approxi-mately constant with radius. Inside 18 kpc, the completenessdrops with decreasing radius because of higher stellar crowdingcloser to the nucleus. The total profile is ∼
50% complete at 11kpc.
Figure 13.
Completeness-corrected star count profiles. Verticalerror bars include Poisson noise and completeness uncertainty.The background level and 1 σ uncertainty for each profile, esti-mated from the last 5 bins, are marked as short horizontal linesat right. The RSG, MS+BHeB, and AGB profiles quickly reachthe background levels at R dp ∼
20 kpc. The metal-poor RGB pro-file exhibits a slower decline to the background level compared tothe other profiles. c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 14.
Background CMD covering R dp > . ∼
7% the total field of view. These objects are mostly unre-solved background galaxies and MW foreground stars. The se-lection boxes avoid the most heavily contaminated regions at( V − I ) ∼ − I &
25. The number counts of objects inthe boxes are used to subtract contaminants from the star countprofiles.
The short horizontal lines on the right-hand side ofFig. 13 mark the background levels and 1 σ uncertainties.The background level for each box comes from the totalcounts and area summed over the last 5 bins covering R dp > . ∼
130 arcmin or ∼
7% the total field of view. There are no obvious NGC 2403stellar sequences visible, consistent with these objects be-ing dominated by unresolved background galaxies and MWforeground stars. The selection boxes avoid the most heavilycontaminated regions at ( V − I ) ∼ − I & R dp ∼
18 kpc. There isalso a small excess of metal-rich RGB stars at 18 −
27 kpc,but this could be due to photometric errors scattering somemetal-poor RGB stars into the metal-rich box.The MS+BHeB and RSG profiles extend out to ∼ . R and the AGB profile extends out to R dp = 27kpc or 2 . R , consistent with the findings of Davidge (2003,2007). The AGB profile shows a slight change in slope at 18kpc similar to the RGB profiles, but the 1 σ errors on thepoints beyond this distance are too large to say with highconfidence whether or not this change is real.All the background-subtracted profiles exhibit a similar Figure 15.
Background-subtracted, completeness-corrected starcount profiles for the CMD selection boxes. The background levelsand 1 σ uncertainties are marked by the short horizontal lines atright. There is an excess of metal-poor RGB stars relative to theextrapolation of the inner profile outside the main disc, indicatingthe presence of a metal-poor extended structure. The errors barsinclude the uncertainty in the background level. Table 2.
Radial scale-lengths for star count profiles in therange R dp = 9 −
17 kpc.CMD box Scale-length(kpc)MS+BHeB 1 . ± . . ± . . ± . . ± . . ± . . ± . . ± . steep slope at R dp ∼ −
17 kpc. As we will see in Sec-tion 5, this radial range is clearly dominated by outer disc.We fitted exponentials to the profiles in this region, aftermultiplying the logarithm of the surface density by 2.5 tobring them onto a magnitude scale, and the resulting scale-lengths are listed in Table 2. The quoted errors give theinterval over which χ ν increases by 1.0. The young star pro-files have scale-lengths around 1.7 kpc while the AGB starshave a larger scale-length of ∼ . c (cid:13) , 1–20 Barker et al.
Figure 16.
Colour histogram for all point sources with I =23 . − . V − I ) = 1 . − . To examine in more detail the nature of the extended RGBcomponent at large radii, Fig. 16 shows the colour histogramfor point sources in the magnitude range I = 23 . − . R dp = 40 . − . V − I ) ∼ . − .
4, whichdrops to the background level at the outermost radii. Thereis little apparent change in the position of the peak withradius suggesting little variation in the peak metallicity. At R dp = 27 . − . α/F e ] = 0and adopt the Marigo et al. (2008) isochrones. We use 9isochrones spaced roughly 0.3 dex apart in the range [M/H]= –2.3 to 0.2. These isochrones form an irregular grid incolour, magnitude, and metallicity. We then interpolate be-tween the grid points to measure the metallicity of eachsource in the RGB selection box. The interpolation is per-formed with the TRIGRID function in the Interactive Data Figure 17.
RGB MDF for several different radial ranges usingthe Marigo et al. (2008) isochrones and assuming an age of 10 Gyrand [ α/F e ] = 0. Using the Dotter et al. (2007) isochrones shiftsthe peak ∼ . Language (IDL), which utilizes Delaunay triangulation andpolynomial interpolation.Fig. 17 shows the MDFs for different radial ranges inunits of surface density per 0.2 dex-wide bin. Sources fallingin the metal-rich RGB box lie within the 2 highest metal-licity bins. We do not show metallicities higher than –0.4 asthis is roughly the maximum metallicity contained withinthe metal-rich RGB box for this age. The MDFs are not cor-rected for incompleteness, but this should not dramaticallyalter their shapes since the metal-rich RGB box is typicallyonly ∼
5% less complete than the metal-poor RGB box.The MDFs for all radii exhibit a peak at [M/H] = − . ∼ . − . ± . ∼ . α -elements would leave our estimate for [M/H]unchanged because α -enhanced isochrones can be approxi-mated by scaled-solar isochrones with the same global metal-licity. Using the formalism of Salaris et al. (1993), an en-hancement in the α -elements of [ α/F e ] = 0 . ∼ . c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 18.
Comparing ACS data to Subaru data for fields H2+H6 ( R dp ∼
11 and 12 kpc) and H3+H7 ( R dp ∼
16 and 21 kpc). The leftcolumn compares the CMDs for all point sources in the ACS and Subaru catalogues and Marigo et al. (2008) isochrones for an age of10 Gyr and [M/H] = –1.3, –0.7, and –0.4. The right column shows the MDFs for all RGB stars and for the matched subsets in the twophotometric metallicity systems. The MDF peaks in the two systems agree to within ∼ . stars decreases, the MDFs look increasingly similar to thebackground. At all radii, there appears to be a large spreadin metallicity. The artificial star tests indicate that only at R dp .
18 kpc do the photometric errors contribute signif-icantly to this spread. At these radii, there could also besome young, red helium-burning giants contaminating themetal-poor tail. A spread in age at any radius may also con-tribute to the observed MDF widths. Finally, we note that astar-by-star scatter plot of [M/H] versus R dp did not revealany further information.Davidge (2003) observed a single 5 . ′ × . ′ field locatedon the NE minor axis at R dp ≈ . / H] = − . ± . ± . R dp ∼
15 kpcand [Fe / H] = − . ± . ± . R dp ∼ following the steps outlinedin the DOLPHOT manual and using the default input pa-rameters. We defined objects as point sources if they wereclassified by DOLPHOT as ’good stars’ with S/N > < . DOLPHOT is an adaptation of the photometry pack-age HSTphot (Dolphin 2000). It can be downloaded fromhttp://purcell.as.arizona.edu/dolphot/.c (cid:13) , 1–20 Barker et al. | sharp | < . χ <
3. Magnitudes were reported for everysource in the native ACS VEGAMAG filter system ( F W and F W ) and in the ground-based Johnson-Cousins sys-tem using the transformation equations in Sirianni et al.(2005).Fig. 18 compares the ACS CMDs and MDFs with theSubaru CMDs and MDFs for all point sources falling withinthe ACS fields. To boost the number statistics, we combinedthe ACS fields into two pairs with similar radii, R dp ∼ R dp ∼
16 and 21 kpc for H3+H7.Overplotted on the CMDs are the Marigo et al. (2008)isochrones for an age of 10 Gyr and [M/H] = –1.3, –0.7,and –0.4. For ease of comparison, the ACS CMDs are in theground-based magnitude system, but to avoid any possiblebiases introduced in the transformation, the computation ofthe ACS MDFs was done in the native ACS system. Addi-tionally, we matched point sources in the Subaru cataloguewith point sources in the ACS catalogues by applying con-stant offsets of ∼ . ′′ − . ′′ in right ascension and declina-tion. The MDFs for the matched subsets are also shown inFig. 18.Overall, there is good correspondence between the ACSand Subaru CMDs. The RGB is clearly visible in both CMDsdespite the higher background contamination in the ground-based data which dominates in H3+H7 at 0 . ( V − I ) . V − I ) ∼ Having established the existence of an extended structuralcomponent at large radii in NGC 2403, we now construct acomposite SB profile which uses diffuse light and resolvedstar counts in the regions where they are each most reliable.Within the bright optical disc, where the effects of incom-pleteness on the star counts are most severe, we use thediffuse light because of its insensitivity to these effects. Inthe outer regions, where the sky background dominates thediffuse light, but where the completeness rate is the highestand varies the least, we use the total star counts, which havea higher contrast over the background than the diffuse light.We derived the V-band diffuse light SB profile ofNGC 2403 using the IRAF ellipse task with elliptical an-nuli of constant center, position angle, and inclination af-ter masking saturated stars. In each elliptical annulus, themedian pixel value was computed after two 5 σ clipping it-erations. The diamonds in Fig. 19 show the sky-subtractedprofile for the average of both fields while the circles showthe profile for each field separately. The sky value for eachfield was estimated from the mode of the pixel histogram,which was 1930 ADU and 2660 ADU in F1 and F2, re-spectively. These values translate to a sky SB of µ V =21 . − and µ V = 21 . − . Anothermethod of sky estimation, that involved taking the mean ofthe median pixel value in 16 3 ′ x3 ′ boxes near the edges of the Figure 19.
Diffuse light SB profile for F1 (solid circles) F2 (opencircles), and the average of both fields (diamonds). Error bars areshown only for the average profile and they include read noise,sky background uncertainty, and the r.m.s. variation within eachannulus. The best-fit exponential disc (solid line) has a scale-length h = 2 . ± . mosaic, gave values 10 ADU and 18 ADU less than the firstmethod, or about 0.9 and 1.5 times the standard deviationof the median pixel box values in F1 and F2, respectively.To be conservative, we adopt the difference between the twomethods as the sky uncertainty. The error bars are shownonly for the average profile and they include read noise, skybackground uncertainty, and the root-mean-square (r.m.s.)deviation of the pixel values in each annulus to account forazimuthal variations due to spiral arms, HII regions, OB as-sociations, and any possible warping of the stellar disc. Themedian foreground extinction of all point sources, A V = 0 . h = 2 . ± . ∼ h = 1 . ± . h = 1 . µ ( R ) = − . [Σ(R)] + ZP, where Σ( R ) isthe stellar surface density and the zero-point, ZP = 31.32,is estimated from the overlapping region between the starcounts and diffuse light. By merging the profiles in this way,we can trace NGC 2403’s SB over a larger radial range thanis possible with either profile alone. This is made possible bythe much lower sky background attained with the resolvedstar counts, µ V ∼
29 mag arcsec − , more than 7 magnitudesfainter than the diffuse light background. c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 20.
Composite SB profile for NGC 2403 made from thediffuse light (diamonds) and total star count profiles (squares).The composite profile traces the SB of NGC 2403 down to µ V ∼
32 mag arcsec − . Figure 21.
Two different disc+halo decompositions of the SBprofile: (1) exponential disc with scale-length h = 2 . ± .
03 kpcand Hernquist halo with scale-radius fixed at the MW’s value r s = 14 . h = 2 . ± .
03 kpcand Hernquist halo with r s = 1 . +1 . − . kpc. The points and errorbars are the same as in Fig. 20. Solid lines show the total modelprofiles and broken lines show individual components. Figure 22.
Two different disc+disc decompositions of the SBprofile: (1) inner disc exponential scale-length h = 2 . ± . h = 3 . h = 2 . ± .
03 kpcand outer disc scale-length h = 15 ± In Fig. 20, we show the composite SB profile forNGC 2403 made by combining the diffuse light (diamonds)and total star count profiles (squares). The SB profile ex-tends over 12 magnitudes and reaches µ V ∼
31 mag arcsec − at R dp ∼
30 kpc and µ V ∼
32 mag arcsec − at R dp ∼ h = 8 . ± . I ( r ) ∝ r − γ ) with index γ = 3 . ± . V -bandluminosities of their constituent stars before summing themto make the total profile. We can estimate the magnitudeof this effect by scaling the star count profiles by the aver-age V-band luminosities of point sources in their respectiveCMD boxes. This has the effect of shifting the young starand AGB profiles upward relative to the RGB profiles andslightly steepening the total star count profile scale-lengthin the region R dp = 9 −
17 kpc from 2.2 kpc to 2.0 kpc.Thus, we expect that disregarding this effect will have lit-tle impact on our results. We also note that correcting forinternal extinction in the main gas disc could steepen theprofile, as well, but we have chosen not to attempt any cor-rection for this because of the uncertain dust properties inNGC 2403.Fig. 21 shows two separate models for NGC 2403’s c (cid:13) , 1–20 Barker et al.
SB profile which include an exponential inner disc anda spherically-symmetric Hernquist halo. We recall that aHernquist profile can be characterised by a scale radius, r s ,which is approximately 41% of the half-mass radius andthat, for r >> r s , the projected light profile follows apower-law with exponent γ = 3 (Hernquist 1990). In thefirst model, the best-fit exponential inner disc scale-lengthis h = 2 . ± .
03 kpc and the halo’s scale radius is heldfixed at r s = 14 . −
20 kpcfound in some theoretical semi-analytic simulations (Bullock& Johnston 2005), and it provides a reasonable descriptionof M81’s extended component (Barker et al. 2009). If we al-low the halo’s scale radius to be free, then its best-fit valueis r s = 1 . +1 . − . kpc and the disc scale-length is virtually un-changed at h = 2 . ± .
03 kpc. Both values of r s providesimilar fit qualities and yield similar halo luminosities. If weextrapolate the fits out to 50 kpc, then the haloes wouldcontain ∼ −
5% of the total galactic V-band luminosity,or L V ∼ − × L ⊙ . Note that the luminosity does notsignificantly change if extrapolated further out to 100 kpc.If the extended component is a disc structure, then itwould be more appropriate to describe it with a radial expo-nential profile than a Hernquist profile. We show two differ-ent disc plus disc models of the SB profile in Fig. 22. In thefirst, the scale-length of the extended component is fixed tothat of the MW’s thick disc, h = 3 . h = 2 . ± .
03 kpc. However inspection of Fig. 22 showsthis model clearly fails to explain the excess light beyond20 kpc. Thus, the scale-length of the extended componentcould only be as small as the MW’s thick disc if we havegrossly underestimated the background or if there is a thirdstructural component that dominates at R dp &
30 kpc. Inthe second model, the extended component’s scale-length isfree and its best-fit value is h = 15 ± h = 2 . ± .
03 kpc. The V-bandluminosity of the outer disc in these two models is ∼ − V -bandluminosity is somewhat dependent on the model adopted,particularly on the behaviour of the model inside 20 kpc. Ifthis component exists at all radii, then the range ∼ − × L ⊙ ( ∼ −
7% of the total) is likely to bracket the trueluminosity.
We have found strong evidence for an extended structure ofRGB stars in NGC 2403 which dominates the light profilebeyond R dp ∼
18 kpc and has a peak metallicity of [M/H]= − . ± .
3. This structure has a flatter radial profile thanthe inner disc, can be reliably traced to R dp ∼
40 kpc and µ V ∼
32 mag arcsec − . The radial profile is consistent with a power-law halo or exponential disc and we now discussthese possibilities in more detail.The MW and M31 have the most thoroughly studiedstellar haloes, so they provide a useful baseline for compari-son to the extended component in NGC 2403. Although theMW’s stellar halo contains significant amounts of substruc-ture (e.g. Bell et al. 2008), many studies over the years havefound that it can be broadly described by a power-law vol-ume density distribution with index Γ = γ + 1 ∼ r s ≈
14 kpc. The total luminosityof the MW halo was estimated by Carney et al. (1990) to be L V ∼ L ⊙ , which is ∼
5% of the total galactic luminosityof ∼ × L ⊙ (Sackett 1997).Accretion remnants are also observed throughout thehalo of M31, but Ibata et al. (2007) found that a subregion ofthe southern quadrant lacked any substructure and could befit out to 150 kpc by a Hernquist profile with r s = 53 . ± . h = 46 . ± . γ = 1 . ± .
12. They estimated a total halo V -band luminosity similar to the MW’s halo, corresponding to ∼
2% of the total of ∼ × L ⊙ (de Vaucouleurs et al.1991).In their analytic simulations of hierarchical galaxy for-mation, Purcell et al. (2007) found that the fraction of dif-fuse halo light increases with host galaxy total mass. Thefractional luminosity of the extended component in NGC2403 is of the same order as the MW and M31 stellar haloes,and higher than the expected value of ∼ .
3% for a 10 M ⊙ dark matter (DM) halo in the Purcell et al. (2007) simula-tions. Their simulations had considerable scatter at fixedhost galaxy mass reflecting variations in the mass accretionhistory. Indeed, a fractional luminosity of 1%, on the lowend of our estimates, is within the 95% confidence intervalfor their simulations of haloes with this mass. Given the un-certainties in the mass of NGC 2403’s DM halo and the starformation prescriptions employed in the simulations, thesedifferences are not entirely surprising and cannot be used toexclude the idea that the extended component is a stellarhalo. What is more clear is that the extended component inNGC 2403 is less luminous than the MW and M31 haloes,as would be expected if stellar halo luminosity scales withtotal galaxy luminosity.The MW halo has a peak metallicity [Fe/H] ∼ − .
6, or[M/H] ∼ − . α/ Fe] ∼ . c (cid:13) , 1–20 he Faint Structure of NGC 2403 Figure 23.
Major and minor axis SB profiles (upper and lowerpoints, respectively) constructed using circular annuli in 40 ◦ -widewedges centred on the axes. The diamonds and squares show thediffuse light and total star count profiles, respectively. The pointsare colour-coded so that the north-west major axis is black, south-east major axis is blue, north-east minor axis is red, and south-west minor axis is orange. The black lines show the disc+halomodels from Fig. 21 with halo scale radius r s = 14 . h = 3 . predict a spread in halo properties at a given mass depend-ing on the exact details of the accretion history.The concept of a single well-defined stellar halo in theMW may not be totally accurate. Carollo et al. (2010) foundevidence that the MW halo could be divided into two kine-matically distinct components, with a flattened inner halodominating at ∼ −
10 kpc and a roughly spherical outerhalo dominating beyond 20 kpc. The inner halo density dis-tribution had a power-law index Γ = 3 . ± . − . . ± .
29 and a peak [Fe/H] = − .
2. These resultshighlight the fact that the inferred properties can depend onthe region observed. If the NGC 2403 extended component isa dual-component halo with the same radial divisions as theMW halo, then, because the inner disc dominates the lightat small radii, our observations would be most sensitive tothe transition zone between the inner and outer haloes andthe beginning of the outer halo.To explore further if the extended component could be aspherically symmetric halo, Fig. 23 plots the SB profile alongthe major and minor axes (upper and lower points, respec-tively). These profiles were constructed using circular an-nuli in 40 ◦ -wide wedges centred on each axis. The diamondsand squares show the diffuse light and total star count pro-files, respectively. The points are colour-coded so that thenorth-west major axis is black, south-east major axis is blue,north-east minor axis is red, and south-west minor axis is orange. The black lines show the disc+halo models from § r s = 14 . § h = 3 . R pr ∼
17 kpc. The disc+discmodel with outer disc scale-length h = 15 kpc provides thebest fit to the major axis profiles. Thus, the data show aclear preference for a flattened geometry for the extendedcomponent with a projected axis ratio similar to that of theinner disc.If it is a disc structure, the extended component may beanalagous to the thick discs observed in the MW and edge-on galaxies. Since we lack a full understanding of how thickdiscs form, it is unclear exactly how their stellar populationsshould scale with host galaxy mass. The scale-length and lu-minosity of the extended component in NGC 2403 are longerand smaller, respectively, than those of the MW’s thick disc,which has a radial scale-length of ∼ − ∼
15% to theMW’s total disc luminosity (Buser et al. 1999; Chen et al.2001; Larsen & Humphreys 2003). The extended compo-nent’s scale-length is also larger than that of the extendeddisc and the thick disc in M31, which are h = 5 . ± . h = 8 . ± . ∼ − . ∼ − . − . ∼ −
300 Myr ago (Yun et al. 1994; Yun 1999).Barker et al. (2009) identified an extended stellar compo-nent around M81 that could be its halo or an extended disclike that in M31. This component bears some striking dif- c (cid:13) , 1–20 Barker et al. ferences to the extended component in NGC 2403. Disre-garding any contribution from the inner disc, the total starcounts beyond R dp ∼
20 kpc had a radial scale-length of h = 12 . ± . γ = 2 . ± . µ V ∼
26 mag arcsec − and the implied luminosity was L V ∼ − × L ⊙ if extrapolated out to 100 kpc overall position angles. There was an RGB visible in the CMDat R dp = 32 −
44 kpc, which was estimated to have a peakmetallicity [M/H] = − . ± . ∼ . − .The azimuthal star count profile for the metal-poor RGBwas somewhat flatter than that for the metal-rich RGB andAGB, suggesting that the extended component was a haloor a more face-on or thicker disc structure.In contrast, the extended component in NGC 2403starts to dominate over the disc at a lower SB and theimplied luminosity is ∼
10 times less. The radial profile issteeper and there are no detectable AGB or metal-rich RGBstars beyond 30 kpc. The axis ratio of the extended compo-nent does not appear significantly different from that of theinner disc out to R pr ∼
25 kpc.The few similarities that do exist between the M81 andNGC 2403 extended components include the fact that bothstart to dominate over the bright optical disc at R dp ∼ r / -law profile couldnot fit both M81’s bulge and its extended component.It is interesting, then, that NGC 2403 so far shows noclear-cut signs of interaction in its stellar or HI distribu-tions yet it still has an extended component that could bea thick disc or halo. Our survey has found no obvious stel-lar streams or substructures around NGC 2403 that wouldattest to a significant recent accretion, but structures likethe extended tails of RGB stars observed around M33 (Mc-Connachie et al. 2010) would be too faint for us to observe.Regions beyond the area we have surveyed are more likelyto contain substructures because of the longer phase mixingtimescale there (Johnston et al. 2008).Finally, we note that the main (inner) exponential discof NGC 2403 dominates the surface brightness profile to ∼
18 kpc, or & ∼
10 scalelengths (15 kpc) where it hasa peak metallicity of [Fe/H] = –0.9. This finding providesfurther evidence that the stellar discs of spiral galaxies canoften be far more extended than commonly thought.
Using Suprime-Cam on the Subaru telescope we have con-ducted a wide-field imaging survey of RGB stars around thelow mass spiral galaxy NGC 2403. These observations rep-resent the first global analysis of RGB stars in a late-typespiral beyond the Local Group. The surveyed area reaches a maximum R pr ∼
30 kpc or R dp ∼
60 kpc. The CMD reaches1.5 mag below the tip of the metal-poor RGB at a complete-ness rate >
50% for R dp &
12 kpc. We detect young stars(ages ∼ −
200 Myr) out to radii ∼ . R , or R dp ∼ & R dp ∼
18 kpc, and reaches a V -band SB of µ V ∼
29 mag arcsec − . Beyond this radius, we find strongevidence for an extended structural component with a flatterSB profile than the inner disc and which we trace out to R dp ∼
40 kpc and µ V ∼
32 mag arcsec − . The extendedcomponent’s V-band luminosity integrated over all radii isone to a few percent that of the whole galaxy, depending onassumed profile. At R dp ∼ −
30 kpc, we estimate a peakmetallicity [M/H] = − . ± . α/ Fe] = 0. The projected axis ratio of the extendedcomponent does not appear significantly different from thatof the inner disc within R pr ∼
25 kpc.Possible interpretations for the nature of this com-ponent include a inestellar halo or thick disc. Kinematicinformation for tracer populations would help distinguishwhether this component is a rotating disc structure or apressure-supported halo. There are few, if any, bright AGBstars in this component to act as spectroscopic targets, sowe must wait for the next generation of facilities to targetthe more numerous RGB stars. These results provide fur-ther evidence that faint, extended stellar structures appearto be a generic feature of disc galaxies, even isolated late-type systems.
ACKNOWLEDGMENTS
MKB and AMNF acknowledge support from a Marie CurieExcellence Grant from the European Commission under con-tract MCEXT-CT-2005-025869 and a rolling grant from theScience and Technology Facilities Council.
REFERENCES
Allende Prieto C., Beers T. C., Wilhelm R., Newberg H. J.,Rockosi C. M., Yanny B., Lee Y. S., 2006, ApJ, 636, 804Bailin J., Bell E. F., Chappell S. N., Radburn-Smith D. J.,de Jong R. S., 2011, ApJ, 736, 24Barker M. K., Ferguson A. M. N., Irwin M., Arimoto N.,Jablonka P., 2009, AJ, 138, 1469Bell E. F., Zucker D. B., Belokurov V., Sharma S., JohnstonK. V., Bullock J. S., Hogg D. W., Jahnke K., de JongJ. T. A., Beers T. C., Evans N. W., Grebel E. K., Ivezi´cˇZ., Koposov S. E., Rix H.-W., Schneider D. P., SteinmetzM., Zolotov A., 2008, ApJ, 680, 295Brook C. B., Kawata D., Gibson B. K., Freeman K. C.,2004, ApJ, 612, 894Bullock J. S., Johnston K. V., 2005, ApJ, 635, 931Burstein D., 1979, ApJ, 234, 829Buser R., Rong J., Karaali S., 1999, A&A, 348, 98 c (cid:13) , 1–20 he Faint Structure of NGC 2403 Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345,245Carney B. W., Latham D. W., Laird J. B., 1990, AJ, 99,572Carollo D., Beers T. C., Chiba M., Norris J. E., FreemanK. C., Lee Y. S., Ivezi´c ˇZ., Rockosi C. M., Yanny B., 2010,ApJ, 712, 692Chapman S. C., Ibata R., Lewis G. F., Ferguson A. M. N.,Irwin M., McConnachie A., Tanvir N., 2006, ApJ, 653,255Chen B., Stoughton C., Smith J. A., Uomoto A., Pier J. R.,Yanny B., Ivezi´c ˇZ., York D. G., Anderson J. E., AnnisJ., Brinkmann J., Csabai I., Fukugita M., Hindsley R.,Lupton R., Munn J. A., the SDSS Collaboration 2001,ApJ, 553, 184Chynoweth K. M., Langston G. I., Holley-Bockelmann K.,Lockman F. J., 2009, AJ, 138, 287Collins M. L. M., Chapman S. C., Ibata R. A., Irwin M. J.,Rich R. M., Ferguson A. M. N., Lewis G. F., Tanvir N.,Koch A., 2011, MNRAS, 413, 1548Cooper A. P., Cole S., Frenk C. S., White S. D. M., Helly J.,Benson A. J., De Lucia G., Helmi A., Jenkins A., NavarroJ. F., Springel V., Wang J., 2010, MNRAS, 406, 744Cutri R. M. e. a., 2003, NASA/IPAC Infrared Sci-ence Archive, 2MASS All Sky Catalog of point sources,http://irsa.ipac.caltech.edu/applications/Gator/Dalcanton J. J., Bernstein R. A., 2002, AJ, 124, 1328Dalcanton J. J., Williams B. F., Seth A. C., Dolphin A.,Holtzman J., Rosema K., Skillman E. D., Cole A., GirardiL., Gogarten S. M., 2009, ApJS, 183, 67Davidge T. J., 2003, AJ, 125, 3046Davidge T. J., 2007, ApJ, 664, 820de Grijs R., van der Kruit P. C., 1996, A&AS, 117, 19de Jong R. S., 2008, MNRASde Jong R. S., Radburn-Smith D. J., Sick J. N., 2008, inASP Conf. Ser. 396 ed. J. G. Funes & E. M. Corsini (SanFrancisco, CA: ASP), GHOSTS — Bulges, Halos, andthe Resolved Stellar Outskirts of Massive Disk Galaxies.p. 187de Vaucouleurs G., de Vaucouleurs A., Corwin Jr. H. G.,Buta R. J., Paturel G., Fouque P., 1991, Third ReferenceCatalogue of Bright Galaxies. New York: SpringerDolphin A. E., 2000, PASP, 112, 1383Dotter A., Chaboyer B., Jevremovi´c D., Baron E., FergusonJ. W., Sarajedini A., Anderson J., 2007, AJ, 134, 376Ferguson A., Irwin M., Chapman S., Ibata R., Lewis G.,Tanvir N., 2007, in Island Universes: Structure and Evo-lution of Disk Galaxies, ed. R. S. de Jong (Dordrecht:Springer), p. 239Ferguson A. M. N., Irwin M. J., Ibata R. A., Lewis G. F.,Tanvir N. R., 2002, AJ, 124, 1452Font A. S., Johnston K. V., Bullock J. S., Robertson B. E.,2006, ApJ, 646, 886Fraternali F., van Moorsel G., Sancisi R., Oosterloo T.,2002, AJ, 123, 3124Freedman W. L., Madore B. F., Gibson B. K., Ferrarese L.,Kelson D. D., Sakai S., Mould J. R., Kennicutt Jr. R. C.,Ford H. C., Graham J. A., Huchra J. P., Hughes S. M. G.,Illingworth G. D., Macri L. M., Stetson P. B., 2001, ApJ,553, 47Gallart C., Zoccali M., Aparicio A., 2005, ARA&A, 43, 387Garnett D. R., Shields G. A., Skillman E. D., Sagan S. P., Dufour R. J., 1997, ApJ, 489, 63Geisler D., Wallerstein G., Smith V. V., Casetti-DinescuD. I., 2007, PASP, 119, 939Gilmore G., Wyse R. F. G., Jones J. B., 1995, AJ, 109,1095Greisen E. W., Calabretta M. R., 2002, A&A, 395, 1061Helmi A., 2008, A&A Rev., 15, 145Hernquist L., 1990, ApJ, 356, 359Ibata R., Chapman S., Ferguson A. M. N., Lewis G., IrwinM., Tanvir N., 2005, ApJ, 634, 287Ibata R., Martin N. F., Irwin M., Chapman S., FergusonA. M. N., Lewis G. F., McConnachie A. W., 2007, ApJ,671, 1591Ibata R., Mouhcine M., Rejkuba M., 2009, MNRAS, 395,126Irwin M., Lewis J., 2001, NewA Rev., 45, 105Irwin M. J., 1985, MNRAS, 214, 575Irwin M. J., 1997, in Espinoza J. M., ed., 7th Canary Is-lands Winter School Detectors and Data Analysis Tech-niques for Wide Field Optical Imaging. Cambridge: Cam-bridge Univ. Pr., p. 35Irwin M. J., Lewis J., Hodgkin S., Bunclark P., EvansD., McMahon R., Emerson J. P., Stewart M., Beard S.,2004, in Quinn P. J., Bridger A., eds, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Se-ries Vol. 5493 of Presented at the Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference, VISTAdata flow system: pipeline processing for WFCAM andVISTA. pp 411–422Jablonka P., Tafelmeyer M., Courbin F., FergusonA. M. N., 2010, A&A, 513, A78+Jarrett T. H., Chester T., Cutri R., Schneider S. E., HuchraJ. P., 2003, AJ, 125, 525Johnston K. V., Bullock J. S., Sharma S., Font A., Robert-son B. E., Leitner S. N., 2008, ApJ, 689, 936Juri´c M., Ivezi´c ˇZ., Brooks A., Lupton R. H., Schlegel D.,Finkbeiner D., Krzesinski J., Long D., Neilsen Jr. E. H.,Nitta A., Snedden S. A., York D. G., 2008, ApJ, 673, 864Kalirai J. S., Gilbert K. M., Guhathakurta P., MajewskiS. R., Ostheimer J. C., Rich R. M., Cooper M. C., ReitzelD. B., Patterson R. J., 2006, ApJ, 648, 389Karachentsev I. D., Dolphin A. E., Geisler D., Grebel E. K.,Guhathakurta P., Hodge P. W., Karachentseva V. E.,Sarajedini A., Seitzer P., Sharina M. E., 2002, A&A, 383,125Larsen J. A., Humphreys R. M., 2003, AJ, 125, 1958Loebman S. R., Roskar R., Debattista V. P., Ivezic Z.,Quinn T. R., Wadsley J., 2010, ArXiv e-printsLupton R., Blanton M. R., Fekete G., Hogg D. W.,O’Mullane W., Szalay A., Wherry N., 2004, PASP, 116,133Malin D., Hadley B., 1997, PASA, 14, 52Marigo P., Girardi L., Bressan A., Groenewegen M. A. T.,Silva L., Granato G. L., 2008, A&A, 482, 883Mart´ınez-Delgado D., Aparicio A., Gallart C., 1999, AJ,118, 2229Mart´ınez-Delgado D., Pohlen M., Gabany R. J., MajewskiS. R., Pe˜narrubia J., Palma C., 2009, ApJ, 692, 955McConnachie A. W., 2009, Nature, 461, 66McConnachie A. W., Ferguson A. M. N., Irwin M. J., Du-binski J., Widrow L. M., Dotter A., Ibata R., Lewis G. F.,2010, ApJ, 723, 1038 c (cid:13) , 1–20 Barker et al.
Miyazaki S., Komiyama Y., Sekiguchi M., Okamura S., DoiM., Furusawa H., Hamabe M., Imi K., Kimura M., NakataF., Okada N., Ouchi M., Shimasaku K., Yagi M., YasudaN., 2002, PASJ, 54, 833Morrison H. L., Miller E. D., Harding P., Stinebring D. R.,Boroson T. A., 1997, AJ, 113, 2061Mouhcine M., Ibata R., Rejkuba M., 2010, ApJ, 714, L12Mould J., 2005, AJ, 129, 698Naylor T., 1998, MNRAS, 296, 339Neeser M. J., Sackett P. D., De Marchi G., Paresce F.,2002, A&A, 383, 472Newberg H. J., Yanny B., 2006, J. Phys.: Conf. Ser., 47,195Okamura S., Takase B., Kodaira K., 1977, PASJ, 29, 567Pe˜narrubia J., McConnachie A., Babul A., 2006, ApJ, 650,L33Purcell C. W., Bullock J. S., Zentner A. R., 2007, ApJ,666, 20Purcell C. W., Bullock J. S., Zentner A. R., 2008, MNRAS,391, 550Reitzel D. B., Guhathakurta P., 2002, AJ, 124, 234Rejkuba M., Mouhcine M., Ibata R., 2009, MNRAS, 396,1231Richardson J. C., Ferguson A. M. N., Johnson R. A., IrwinM. J., Tanvir N. R., Faria D. C., Ibata R. A., JohnstonK. V., Lewis G. F., McConnachie A. W., Chapman S. C.,2008, AJ, 135, 1998Richardson J. C., Ferguson A. M. N., Mackey A. D., IrwinM. J., Chapman S. C., Huxor A., Ibata R. A., Lewis G. F.,Tanvir N. R., 2009, MNRAS, 396, 1842Robin A. C., Haywood M., Creze M., Ojha D. K., BienaymeO., 1996, A&A, 305, 125Robin A. C., Reyl´e C., Derri`ere S., Picaud S., 2003, A&A,409, 523Roˇskar R., Debattista V. P., Brooks A. M., Quinn T. R.,Brook C. B., Governato F., Dalcanton J. J., Wadsley J.,2010, MNRAS, 408, 783Sackett P. D., 1997, ApJ, 483, 103Salaris M., Chieffi A., Straniero O., 1993, ApJ, 414, 580Schlegel D. J., Finkbeiner D. P., Davis M., 1998, ApJ, 500,525Sch¨onrich R., Binney J., 2009, MNRAS, 399, 1145Seth A. C., Dalcanton J. J., de Jong R. S., 2005, AJ, 130,1574Shang Z., Zheng Z., Brinks E., Chen J., Burstein D., SuH., Byun Y., Deng L., Deng Z., Fan X., Jiang Z., Li Y.,Lin W., Ma F., Sun W., Wills B., Windhorst R. A., WuH., Xia X., Xu W., Xue S., Yan H., Zhou X., Zhu J., ZouZ., 1998, ApJ, 504, L23+Shao Z., Xiao Q., Shen S., Mo H. J., Xia X., Deng Z., 2007,ApJ, 659, 1159Shaw M. A., Gilmore G., 1990, MNRAS, 242, 59Sirianni M., Jee M. J., Ben´ıtez N., Blakeslee J. P., MartelA. R., Meurer G., Clampin M., De Marchi G., Ford H. C.,Gilliland R., Hartig G. F., Illingworth G. D., Mack J.,McCann W. J., 2005, PASP, 117, 1049Soubiran C., Bienaym´e O., Siebert A., 2003, A&A, 398,141Stetson P. B., 1987, PASP, 99, 191Stetson P. B., 1993, in Butler C. J., Elliott I., eds, IAUColloq. 136, Stellar Photometry - Current Techniques andFuture Developments Further Progress in CCD Photom- etry. pp (Cambridge: Cambridge Univ. Press), 291Stetson P. B., 1994, PASP, 106, 250Stetson P. B., Harris W. E., 1988, AJ, 96, 909Tanaka M., Chiba M., Komiyama Y., Guhathakurta P.,Kalirai J. S., 2011, ArXiv e-printsTikhonov N. A., Galazutdinova O. A., Drozdovsky I. O.,2005, A&A, 431, 127Tsikoudi V., 1979, ApJ, 234, 842Vlaji´c M., Bland-Hawthorn J., Freeman K. C., 2009, ApJ,697, 361Vlaji´c M., Bland-Hawthorn J., Freeman K. C., 2011, ApJ,732, 7Yoachim P., Dalcanton J. J., 2006, AJ, 131, 226Yun M. S., 1999, in Barnes J. E., Sanders D. B., eds, IAUSymp. 186, Galaxy Interactions at Low and High RedshiftTidal Interactions in M81 Group. Paris: IAU, p. 81Yun M. S., Ho P. T. P., Lo K. Y., 1994, Nature, 372, 530Zaritsky D., 1999, AJ, 118, 2824Zibetti S., Ferguson A. M. N., 2004, MNRAS, 352, L6Zibetti S., White S. D. M., Brinkmann J., 2004, MNRAS,347, 556This paper has been typeset from a TEX/ L A TEX file preparedby the author. c (cid:13)000