The Structure of Nuclear Star Clusters in Nearby Late-type Spiral Galaxies from Hubble Space Telescope Wide Field Camera 3 Imaging

2 MB·English

by  Daniel J. Carson

Draftversion January26,2015
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
THE STRUCTURE OF NUCLEAR STAR CLUSTERS IN NEARBY LATE-TYPE SPIRAL GALAXIES FROM
HUBBLE SPACE TELESCOPE WIDE FIELD CAMERA 3 IMAGING *
Daniel J. Carson1, Aaron J. Barth1, Anil C. Seth2, Mark den Brok2, Michele Cappellari3, Jenny E. Greene4,
Luis C. Ho5,6, Nadine Neumayer7
Draft version January 26, 2015
5 ABSTRACT
1 WeobtainedHubbleSpaceTelescope/WideFieldCamera3imagingofasampleoftenofthenearest
0 and brightest nuclear clusters residing in late-type spiral galaxies, in sevenbands that span the near-
2 ultraviolet to the near-infrared. Structural properties of the clusters were measured by fitting two-
dimensional surface brightness profiles to the images using GALFIT.The clusters exhibit a wide range
n
a of structural properties, with F814W absolute magnitudes that range from −11.2 mag to −15.1 mag
J and F814W effective radii that range from 1.4 to 8.3 pc. For six of the ten clusters in our sample, we
findchangesintheeffectiveradiuswithwavelength,suggestingradiallyvaryingstellarpopulations. In
3
2 fouroftheobjects,theeffectiveradiusincreaseswithwavelength,indicatingthepresenceofayounger
population which is more concentrated than the bulk of the stars in the cluster. However, we find a

] general decrease in effective radius with wavelength in two of the objects in our sample, which may
A indicateextended, circumnuclearstarformation. We alsofindageneraltrendofincreasingroundness
G of the clusters at longer wavelengths, as well as a correlation between the axis ratios of the NCs and
their host galaxies. These observations indicate that blue disks aligned with the host galaxy plane
.
h are a common feature of nuclear clusters in late-type galaxies, but are difficult to detect in galaxies
p that are close to face-on. In color-colordiagrams spanning the near-UVthroughthe near-IR,most of
- the clusters lie far from single-burst evolutionary tracks, showing evidence for multi-age populations.
o
Mostoftheclustershaveintegratedcolorsconsistentwithamixofanoldpopulation(>1Gyr)anda
r
t young population (∼100–300Myr). The wide wavelengthcoverageof our data provides a sensitivity
s
topopulationswithamix ofagesthatwouldnotbe possibletoachievewithimaginginopticalbands
a
[ only. The surface brightness profiles presented in this work will be used for future stellar population
modeling and dynamical studies of the clusters.

2
v
6 1. INTRODUCTION ing repeated bursts of star formation, with spectra of-
8 ten consistent with a continuous star formation history
Hubble Space Telescope (HST) imaging surveys have
5 (Walcher et al.2006;Rossa et al.2006;Seth et al.2006).
shown that most galaxies contain bright, compact stel-
5 Young(<100Myr)populationsarefoundinmostspiral
lar systems known as nuclear clusters (NCs) in their
0 NCs, and there is limited evidence for complex popula-
photometric and dynamical centers (Carollo et al. 1997;
1. Matthews et al. 1999; B¨oker et al. 2002). Similar in tions in early-type nuclei as well (Monaco et al. 2009;
0 manyrespectstohigh-massglobularclusters(GCs),NCs Seth et al. 2010; Lyubenova et al. 2013). In a recent
5 typically have velocity dispersions of 15 – 35 km s−1, ef- HST archival study, Georgiev & B¨oker (2014) analyzed
1 the photometric andstructuralpropertiesof a sample of
fective radii (or half-light radii) of a few parsecs, and
: 228NCsinspiralgalaxies,andconfirmedthatmostNCs
v dynamical masses of 106 – 107 M⊙ (B¨oker et al. 1999, are similar in size to GCs, but found that the largest
i 2002; Walcher et al. 2005). Unlike GCs, NCs in spiral
X and brightest NCs in spiral galaxies occupy the regime
galaxieshave complex star formationhistories, undergo-
between ultra-compact dwarfs (UCDs) and the nuclei of
r
a early-type galaxiesin the size-luminosity plane, which is
*Based on observations made with the NASA/ESA Hubble consistent with the idea that some UCDs and GCs are
Space Telescope, obtainedattheSpace TelescopeScienceInsti- the remnant nuclei of tidally disrupted spiral galaxies
tute,whichisoperatedbytheAssociationofUniversitiesforRe-
(B¨oker 2008; Mieske et al. 2013; Seth et al. 2014). They
search inAstronomy, Inc., under NASAcontract NAS5-26555.
TheseobservationsareassociatedwithprogramGO-12163 also found that the effective radius of NCs tends to be
1Department of Physics and Astronomy, University of Cali- smaller when measured in bluer filters, suggesting that
forniaIrvine,4129FrederickReinesHall,Irvine,CA92697,USA youngerstellarpopulationstendtobemoreconcentrated
2Department of Physics and Astronomy, Universityof Utah,
than the rest of the stars in NCs. We note that the
SaltLakeCity,UT84112,USA
3Sub-Department of Astrophysics, Department of Physics, Milky Way’s NC also has a population of young stars
University of Oxford, Denys Wilkinson Building, Keble Road, concentratedin the central0.5 pc (Paumard et al. 2006;
OxfordOX13RH,UK Lu et al. 2013; Scho¨del et al. 2014).
4DepartmentofAstrophysicalSciences,PrincetonUniversity,
Two main scenarios have been proposed to explain
Peyton Hall–IvyLane,Princeton,NJ08544, USA
5KavliInstituteforAstronomyandAstrophysics,PekingUni- the formation of NCs: in-situ formation from the ac-
versity,Beijing100871, China cretion of low angular momentum gas onto the galac-
6Department of Astronomy, School of Physics, Peking Uni-
tic nucleus (Milosavljevi´c2004), andmergersofglobular
versity,Beijing100871, China
7Max-Planck-Institut fu¨r Astronomie, K¨onigstuhl 17, D- clusters migrating towards the galactic nucleus due to
69117Heidelberg,Germany dynamical friction (Lotz et al. 2001). Hartmann et al.
2
(2011) compared simulations of the migratory/merger near-IRIFUspectroscopy,opticalspectroscopy,andHST
scenario with integral field unit (IFU) observations of imaging. Usingsimilarmethods,Hartmann et al.(2011)
NGC4244,andfoundthatin-situformationaccountsfor measured a mass of 1.1×107 M⊙ for the NC in NGC
atleast50%oftheNC’smass. Antonini(2013,2014)and 4244,andfoundthatiftheNC hostsacentralBHatall,
Arca-Sedda & Capuzzo-Dolcetta (2014) demonstrated its mass is less than 1% of the NC mass.
thatthe migratory/mergerscenariopredictscorrelations StudiesofNCshavemainlybeenlimitedbytheangular
betweenthemassesofNCsandglobalpropertiesoftheir resolution of existing telescopes. The next generation of
host galaxies that match observed correlations. The large ground-based telescopes, such as the Thirty Meter
young populations observed in most NCs rule out the Telescope (TMT), will be able to resolve the kinematic
sinking of globular clusters as the sole process for NC structure of stars within the sphere of influence of BHs
formation (i.e., the sinking of younger star clusters or with much lower mass than is currently possible (down
gasaccretionmustplaysomerole). However,ourunder- to ∼ 104 M⊙), and will be crucial for future dynami-
standing of the formation of NCs is incomplete and the cal searches for IMBH within NCs (Do et al. 2014). In
reason for the ubiquity of NCs still remains a mystery. a simulation of observations of NCs with the European
NCs and BHs are known to coexist in galaxies of all Extremely Large Telescope (E-ELT), Gullieuszik et al.
Hubble types across a wide range of masses (Seth et al. (2014) found that observations of NCs with telescopes
2008a),butsomeNCshavenocentralBHdowntohighly thatcanresolveindividualstarsintheclusterwillleadto
constraining limits. The NC in NGC 4935, for exam- greatimprovementsin stellar population and age-dating
ple, contains a NC (Filippenko & Ho 2003) as well as a studies,whicharecurrentlylimitedbythevariousdegen-
Type 1 active galactic nucleus (AGN) powered by ac- eraciesthatcomewithfittingstellarpopulationsynthesis
cretion onto a central BH (Filippenko & Sargent 1989), models to the integrated cluster light. The nearest and
withmassestimatesrangingfrom5×104to3.6×105M⊙ brightest NCs will be prime targets for the next genera-
(Peterson et al. 2005; Edri et al. 2012), while dynamical tion of large ground-based telescopes.
studies of the NC in M33, a galaxy with a similar mass We present a set of uniform, high-signal-to-noise
to NGC 4395, have placed remarkably tight upper lim- (S/N),unsaturatedimagesofNCs withwavelengthcov-
its of 1500 – 3000 M⊙ on the mass of any central BH eragefromthenear-ultraviolet(UV)tothenear-infrared
(Gebhardt et al. 2001; Merritt et al. 2001). These stud- (IR) for a sample of ten of the nearestlate-type galaxies
iesindicatethatatleastsomeNCscanhostacentralBH, hosting NCs. Our data represent a substantial improve-
but the overalloccupation fractionof BHs in NCs is un- ment in spatialresolution, wavelengthcoverage,and ho-
known. The masses of both NCs (Ferrarese et al. 2006; mogeneity over previous imaging studies of NCs. Al-
Wehner & Harris 2006) and BHs (Ferrarese & Merritt though all of the galaxies in our sample have archival
2000; Gebhardt et al. 2000; Ha¨ring & Rix 2004) are cor- HST imagesavailablefrompreviousstudies,thearchival
related with the masses of their host galaxy bulges. data are very heterogeneous, including images that are
Galaxies with mass above 1010 M⊙ tend to host BHs, saturated in the core of the NC and images taken with
whilelessmassivegalaxiestendtohostNCs,withatran- several different cameras. The homogeneity of our data
sitionregionforgalaxieswithmass108 –1010 M⊙ where allows for a more detailed analysis of the structure and
both objects coexist (Graham & Spitler 2009). Interest- stellar populations of NCs than would be possible with
ingly,forgalaxiesknowntocontainbothaNCandacen- archival data alone. Detailed structural information of
tral BH, the ratio (MBH +MNC)/Mbulge (where Mbulge thebestresolvedNCsacrossawiderangeofwavelengths
denotes the mass of the bulge or pseudobulge compo- mayprovidecluesaboutformationmechanismsforNCs,
nent of the host galaxy) shows less scatter than either which are currently poorly understood.
MBH/Mbulge or MNC/Mbulge (Kormendy & Ho 2013), Here, we describe the detailed two-dimensional (2D)
suggesting a link between the building of NCs and BHs. structure of this sample of NCs in a uniform way across
These studies add to a growing body of evidence that UV, optical, and IR bands. Surface brightness profiles,
NCs and BHs are both generic by-products of galaxy effective radii, magnitudes, and colors are presented. In
formation, and the growth mechanisms of NCs and BHs future work,we will describe the stellar populations and
are somehow related (Ferrarese et al. 2006). model the stellar dynamics of the NCs in order to con-
Dynamical studies of NCs in late-type spirals have strain the stellar mass and the mass of any central BH
been proven useful as a method for studying BHs. Pre- in each NC. Surface brightness profiles and stellar mass-
vious studies of IC 342 (B¨oker et al. 1999) and NGC to-light profiles of the NCs will be crucial inputs for dy-
3621 (Barth et al. 2009) have constrained MBH by us- namical models of the clusters.
ing HST imaging and integrated stellar velocity disper- The remainder of this paper is organized as follows.
sionmeasurementsoftheNCinordertocarryoutJeans Section 2 describes the NC sample, the HST observa-
modeling of the clusters. Kormendy et al. (2010) and tions and data reduction. In Section 3, we describe sur-
Neumayer & Walcher (2012) have also performed Jeans face brightness profile fits to the images of the clusters.
modeling of NCs in order to set upper limits the masses General results from these fits, which include structural
of central BHs. Adaptive optics (AO) assisted IFU ob- parametersandmagnitudesoftheNCs,trendsofcluster
servations can resolve the kinematics of the nearest and size and roundness with wavelength, and a comparison
brightestNCsandincreasethesensitivitytodetectionsof ofsimple stellar population models to measuredNC col-
BHs within the clusters. For example, Seth et al.(2010) ors are given in Section 4. A brief discussion of each
placedanupper limitof∼105 M⊙ forthe centralBHin individual cluster is given in Section 5. Finally, Section
NGC 404 from dynamical modeling of the stellar kine- 6 summarizes our most important results and describes
matics of the NC using a combination of AO-assisted how they will applied in future work.
3
2. DATA In the IR channel, we used two medium-band filters,
F127M and F153M, to sample the J and H bands be-
We obtained HST/WFC3 images of ten of the near-
cause the NCs are bright enough to saturate with very
est and brightest nuclear clusters in seven different filter
short exposure times (< 10 seconds) using wide-band
bands, spanning the near-UV to near-IR in wavelength.
filters. The IR channel HgCdTe array has coarser res-
The images were taken overa total of 20 HST orbits for
olution than the UVIS channel CCD, with a scale of
the GO-12163observing program.
0′.′13 pixel−1. The IRSUB256 subarray was used, giving
2.1. Sample Selection a 35′′ ×31′′ field of view for the IR images. We used
SPARS10 sampling sequences with NSAMP=10, with
The galaxies in our sample were selected according to
theexceptionofthetwobrightestNCs,IC342andM33,
the following criteria: Hubble type in the range Sc–Sm,
where RAPID readout mode was used to avoid satura-
distance < 5 Mpc (with one exception), and the pres-
tion.
ence of a nuclear star cluster with magnitude brighter
Throughput curves of the HST/WFC3 filters are
than V = 19 mag. The Hubble type selection criterion
shown in Figure 1. Table 2 lists the central wavelengths
ensures that the galaxies in our sample have little or no
and band widths of the filters as well as the exposure
bulge component. NGC 3621, at a distance of 7 Mpc,
times used. Shorter exposure times were used for the
was included in the sample because it is a rare exam-
twogalaxieswiththe brightestNCs,IC 342andM33,in
ple of a nearby galaxy containing a Type 2 AGN in its
order to avoid saturation. Due to the shorter exposure
nuclear cluster (Satyapal et al. 2007; Barth et al. 2009).
timesandRAPIDsampling,the sequenceof4exposures
Basic properties of the galaxy sample are presented in
was duplicated for the IR observations of IC 342 and
Table 1. The morphological types, B-band magnitudes
M33.
and heliocentric radial velocities listed in this table are
We obtained the reddening E(B−V) for each
from The Third Reference Catalogue of Bright Galaxies
galaxy from the latest Galactic dust maps
(de Vaucouleurs et al. 1991) and were looked up in the
(Schlafly & Finkbeiner 2011), and we calculated the
NASA/IPAC Extragalactic Database 1. For eight of the
extinctionduetoGalacticdustineachbandforallofthe
galaxies in our sample, the distances quoted were mea-
galaxies in the sample using the PyRAF/STSDAS task
sured using Cepheid variables. We used tip of the red
Calcphot (Bushouse & Simon 1994). We applied the
giantbranch(TRGB)distancesforNGC 2976andNGC
appropriate reddening to an F star template spectrum
4244,because Cepheid variable distances were not avail-
using the Cardelli et al. (1989) reddening law. From
able.
simulated HST/WFC3 observations, we obtained the
2.2. Description of Observations magnitudes of the reddened and unreddened template
star in each band, and determined the extinction from
Each cluster was observed in seven filter bands over
the difference between the two. All absolute magnitudes
a total of two orbits. For each galaxy, a series of four
listed in this paper have been corrected for Galactic
exposures was taken in each filter band. These four ex-
extinction and are given in the Vega magnitude system.
posures were offset by fractional-pixel shifts relative to
The conversion between magnitudes in the Vega system
each other using a four-point box dither pattern. The
and magnitudes in the AB system (m − m ) is
AB Vega
final images were constructed by combining these expo-
1.50 mag in F275W, 1.18 mag in F336W, −0.15 mag in
sures, which allowed for the removal of cosmic-ray hits
F438W, 0.42 in F814W, 0.96 in F127M, and 1.25 mag
and bad pixels.
in F153M.
In the UVIS channel, we used the F275W, F336W,
F438W,F547M,andF814Wfilters. We usedF547Min- 2.3. Data Reduction
stead of a wide-band filter for the V band because it
The standard HST calibration pipeline was used for
does not include flux from the [O III], Hα, or Hβ emis-
flat fielding, bias correction and dark current subtrac-
sion lines. The UVIS channel CCD has a pixel scale of
tion. In order to reject cosmic-ray hits and bad pixels
0′.′04 pixel−1. At a distance of 5 Mpc, a pixel on the
and to correct for the geometric distortion introduced
UVIS CCD subtends 0.96 pc. The point spread func-
by the optics of HST and WFC3, the four offset expo-
tions (PSFs) range in full width at half maximum from
sures for each filter were combined into a final image
0′.′092inF275Wto 0′.′196inF153M.These angularsizes
using the PyRAF/STSDAS task AstroDrizzle,using a
correspond to physical sizes of 2.20 pc and 4.70 pc at
square kernel with pixfrac = 1, exposure time weight-
a distance of 5 Mpc. Since NCs typically have effective
ing, and sky subtraction turned off. In addition, as part
radii of a few pc, WFC3 provides sufficient spatial reso-
of the AstroDrizzle process, we resampled the IR im-
lution for studies of their structure. In order to observe ages onto the UVIS grid for a uniform pixel size of 0′.′04
each NC with several filters in just two orbits, it is nec-
in all bands. The resulting images are shown in Figure
essarytoreducereadouttimeandeliminatebufferdump
2. Color composite images of each cluster, along with
overheads. Toaccomplishthis,weusedtheMS512Csub-
images of their host galaxies,are shown in Figure 3.
arrayreadoutmode,whichgivesa20′′×20′′fieldofview.
This field of view is more than wide enoughto cover the 3. SURFACEBRIGHTNESSPROFILEFITTING
entire NC and sample the region where the NC surface
A 2D model of the surface brightness profile was fit
brightness profile merges with that of the host galaxy.
to each image using GALFIT version 3.0.4 (Peng et al.
2002, 2010). For each image, we specified the compo-
1 The NASA/IPAC Extragalactic Database (NED) is operated
nents ofthe model surface brightnessprofile to fit to the
by the Jet Propulsion Laboratory, California Institute of Tech-
nology, under contract with the National Aeronautics and Space image (e.g., S´ersic, exponential) and supplied initial es-
Administration. timates forthe model parameters(e.g.,totalmagnitude,
4
TABLE 1
The GalaxySample
Object D (Mpc) m−M Reference Morphology vr (kms−1) HostGalaxyMB E(B−V)
IC342 3.3±0.3 27.58±0.18 1 SAB(rs)cd 31 −23.99 0.480
M33(NGC598) 0.9±0.1 24.65±0.19 2 SA(s)cd −179 −19.05 0.036
NGC247 3.4±0.3 27.68±0.20 2 SAB(s)d 156 −18.82 0.016
NGC300 2.0±0.3 26.48±0.28 2 SA(s)d 144 −18.04 0.011
NGC2403 3.1±0.2 27.43±0.15 3 SAB(s)cd 133 −19.14 0.034
NGC2976 3.6±0.1* 27.77±0.07 4 SAcpec 1 −17.74 0.064
NGC3621 7.3±0.2 29.30±0.06 3 SA(s)d 730 −20.38 0.068
NGC4244 4.3±0.1* 28.16±0.07 4 SA(s)cd 244 −18.96 0.018
NGC4395 4.3±0.4 28.17±0.18 5 SA(s)m 319 −17.66 0.015
NGC7793 3.4±0.1 27.68±0.05 6 SA(s)d 227 −18.38 0.017
References. —Redshift-independentdistancesanddistancemoduli(m−M)wereobtainedfromthefollowing: (1)
Sahaetal. (2002); (2) Bonoetal. (2010); (3) Sahaetal. (2006); (4) Jacobsetal. (2009); (5) Thimetal. (2004); (6)
Pietrzyn´skietal.(2010). AlldistancesweremeasuredusingCepheidvariablesexceptthosemarkedwithanasterisk(*),
whichareTRGBdistances. TheE(B−V)reddeningduetoGalacticdustforeachgalaxyisfromSchlafly&Finkbeiner
(2011).
TABLE 2
HST/WFC3 Filter Properties& Exposure Times
Central Band Exp. Time IC342Exp. M33Exp.
Filter Channel Wavelength (˚A) Width(˚A) (s) Time(s) Time(s)
F275W UVIS 2710.2 164.51 4×380 4×130 4×130
F336W UVIS 3354.8 158.44 4×290 4×55 4×55
F438W UVIS 4326.5 197.30 4×115 4×35 4×42
F547M UVIS 5447.4 206.22 4×95 4×25 4×90
F814W UVIS 8029.5 663.33 4×75 4×15 4×40
F127M IR 12740.0 249.55 4×60 8×5 8×5
F153M IR 15322.0 378.94 4×60 8×5 8×5
Note. — The central wavelength listed is the “pivot wavelength”, a source-independent
measure of the characteristic wavelength of the filter, and the band width is the “passband
rectangularwidth”,theintegralwithrespecttowavelengthofthethroughputacrossthefilter
passbanddividedbythemaximumthroughput,asdefinedintheWFC3InstrumentHandbook
(Dressel2012).
χ2 difference between the HST image and model im-
age is computed, and minimization of χ2 is achieved
(and thus, best-fitting model parameters are found) us-
ingtheLevenberg-Marquardtdownhill-gradientmethod.
To properly weight each pixel in the χ2 calculation, we
used the sigma image generated internally by GALFIT
based on each input image.
We fit 2D surface brightness profiles to the im-
ages rather than fitting 1D radial profiles to elliptical
isophotes of the images, because it allowed us to more
accurately disentangle the instrumental PSF from the
intrinsic profile of the clusters. The clusters are quite
compact and the fits are therefore very sensitive to the
details of the PSF profile. HST/WFC3 PSFs have non-
axisymmetric features that cannot be described by a 1D
radial profile. In addition, we find that some clusters
have strongly non-axisymmetric features such as flat-
tened disk components, which cannot be modeled ac-
curately in 1D. PSF convolution in 1D is therefore not
a suitable alternative to 2D convolution for the purpose
of extracting the intrinsic 2D structure of the NCs from
Fig.1.— Throughput curves for the seven HST/WFC3 filters
taken from the PyRAF/STSDAS task Calcband. Also shown in the HST images.
grayisamodel spectrum ofa1.4Gyr oldstellar population with
solarmetallicityfromBruzual&Charlot(2003).
3.1. PSF Models
position of the center of the profile). A model image Simulated images of the HST/WFC3 PSFs in each
based on these parameters is produced and then con- filter were generated using version 7.4 of the TinyTim
volved with the corresponding HST/WFC3 PSF. The software package (Krist 1993; Krist et al. 2011), which
5
F275W F336W F438W F547M F814W F127M F153M
IC 342
M33
NGC 247
NGC 300
NGC 2403
NGC 2976
NGC 3621
NGC 4244
NGC 4395
2"
NGC 7793
Fig.2.— Inverted grayscale images of the inner 8′′×8′′ of each galaxy in all seven HST/WFC3 filters, shown with an asinh stretch.
Imagesaredisplayedwiththeirnativeorientation.
6
1 arcmin
IC 342 M33 NGC 247 NGC 300 NGC 2403
NGC 2976 NGC 3621 NGC 4244 NGC 4395 NGC 7793
Fig.3.—ColorcompositeimagesofeachNCanditshostgalaxy. Foreachobject,thetoppanelshowstheentirehostgalaxy,themiddle
panel shows the inner 12′′×12′′ of the galaxy, and the bottom panel shows the inner 2′′×2′′ centered on the NC. Host galaxy images
weretakenfromthefollowingsources: 2MASS(Skrutskieetal.2006,IC342),GALEX(GildePazetal.2007,M33),SDSS(Baillardetal.
2011, NGC 2403, NGC 2976, NGC 4244, and NGC 4395), and CGS (Hoetal. 2011, NGC 247, NGC 300, NGC 3621, and NGC 7793).
For the NC images, the HST/WFC3 F275W, F547M, and F814W images were used to populate the blue, green and red color channels,
respectively. The yellow boxes onthe host galaxyimages show the full20′′×20′′ UVISfieldof view and the scalebar denotes 1′. Allof
theimageshavebeenrotatedsothatnorthisupandareshownwithanasinhstretch.
7
creates a model HST PSF based on the based on the totheF814WimageofNGC2403usinga150×150pixel
instrument, detector chip, detector chip position, and fittingregion,weobtainedamagnitudeof15.48mag,an
filter usedinthe observations. To ensurethat the model effectiveradiusof10.00pixels,andaS´ersicindexof1.51
PSFs were processed in the same way as the HST im- for the NC component. When we performed the same
ages,weproducedfourversionsofeachPSFonasubsam- fit using a 300×300 pixel fitting region, we obtained a
pledgridwithsub-pixeloffsets,usingthesamefour-point magnitudeof15.41mag,aneffective radiusof10.82pix-
box dither pattern as the HST/WFC3 exposures. Each els,and a S´ersicindex of1.64 for the NC component. In
model PSF was convolved with the appropriate charge orderto be ableto comparefitresults indifferentfilters,
diffusion kernelin order to accountfor the effect of elec- itisimportanttouseaconsistentfittingregion. Inorder
trons leaking into neighboring pixels on the CCD. The to optimally sample the host galaxy light, we used the
PSFswerethencombinedandresampledontoafinalgrid largest fitting region size that could be applied to every
withapixelsizeof0′.′04inallbandsusingAstroDrizzle, imagewhilestillexcludingbadpixelsontheedgesofthe
with the settings described in Section 2.3. The resulting image introduced during the image combination step of
model PSFs are shown in Figure 4. AstroDrizzle.
We experimented with different functions available in
3.2. Fitting Method
GALFIT to fit to the WFC3 images. Exponential, Gaus-
The NCs sitatthe centeroftheir hostgalaxies,whose sian, and de Vaucouleurs profiles are not flexible enough
surface brightness profiles can generally be described by to describe the wide range of inner profile slopes seen
an exponential disk model because they have either a in the NCs in our sample. Modified King (Elson 1999),
weakbulgecomponentornobulgecomponentatall. The Nuker (Lauer et al. 1995) and S´ersic (1968) profiles all
disk of the host galaxy contributes a significant amount provided good fits to the F814W images of M33, NGC
of flux in our images, particularly in the outskirts of the 300, and NGC 7793. These NCs have relatively simple
NCs, and can have a strong effect on NC fit parame- structures and serve as instructive cases for comparison
ters such as the S´ersic index and effective radius. Our of different surface brightness models. In each case, the
images have a small field of view and only sample the S´ersicprofileresultsinthe smallestχ2,althoughthe dif-
veryinnerregionsofthegalaxies. Forexample,themost ference in χ2 between fits performed using S´ersic and
distant galaxy in our sample, NGC 3621, has apparent King profiles is always at the 10% level or less. The
angulardimensions12′×7′(Gil de Paz et al.2007),com- Nuker profile has five independent parameters and the
paredto the 20′′×20′′ UVIS field ofview. Although the Kingprofilehasfour,whiletheS´ersicprofileisdescribed
NCs sit against a non-uniform background, it is not al- by only three independent parameters: the total magni-
wayspossibletodetectagradientinthebackgroundlevel tude,effectiveradius(half-lightradius)andS´ersicindex.
across the image due to the small field of view. In order We used a S´ersic profile to model the light profile of the
to model the host galaxy surface brightness profile, we NCs in our sample because the S´ersic profile provides
masked out the NC as well as any prominent dust lanes the best flexibility in describing the structure of differ-
and attempted to fit a model consisting of an exponen- ent NCs using the smallest number of free parameters.
tialdisk anda flatskylevelto the hostgalaxy. Ifthis fit To allow for elliptical shapes we left the axis ratio (b/a,
converged,weperformedtheNCfitwiththehostgalaxy wherebandaarethelengthsofthesemiminorandsemi-
exponentialdisk andskyparametersfixedatthe best-fit major axes, respectively) and the position angle (PA) of
values found in the previous step. We were only able to the semimajor axis as free parameters in the fits.
fit the host galaxy with an exponential disk and flat sky WiththeexceptionofNGC2976,NGC4395andNGC
levelusingthismethodforIC342,NGC3621,andNGC 4244,asingleS´ersicprofilewasusedtomodeleachNCin
4244. For the rest of the sample, the scale length and everyband. InthebandsbluewardofF814W,the struc-
total magnitude were flagged as a problematic parame- ture of the NC in NGC 2976 is difficult to fit with any
ters by GALFIT, indicating that the host galaxy profile reasonable set of analytic models due to its small-scale
is too flat across the image to be modeled by an expo- clumpiness. This NC also contains a compact blue fore-
nentialdiskprofile. ForM33,NGC247,NGC 300,NGC groundcomponenttothenorth(seeFigure3). ForNGC
2403, NGC 2976, NGC 4395 and NGC 7793, we used a 2976,we usedaperture photometry to measurethe total
flat sky level to model the host galaxy light and left its magnitude of the cluster and the magnitude of the blue
surface brightness as a free parameter when fitting the componentinF275W,F336W,andF438W,andF547M.
NC surface brightness profile. Since the blue component becomes faint and the struc-
For each fit, we used a 401× 401 pixel (16′′ × 16′′) tureoftheNC becomessmootheratlongerwavelengths,
fitting region centered on the brightest pixel in the NC. weusedGALFITtofitasingleS´ersicprofiletothecluster
This fitting region size is quite large compared to the inF814W,F127MandF153M.Seth et al.(2008b)found
typical angular size of NCs in our sample, whose effec- thattheNCinNGC4244hastwodistinctmorphological
tive radii are all well under 1′′. We experimented with a components: a red/old, compact, spheroidal structure,
rangeofdifferentfittingregionsizestoquantifytheeffect and a blue/young, extended disk structure, while stud-
of the fitting region size on the fit results and find that iesbyFilippenko & Sargent(1989)andFilippenko & Ho
these effects are at the 10 % level for the nuclear clus- (2003)haveshownthatNGC 4395hostsa NC aswellas
ter properties. For a single S´ersic component fit with a anunobscuredAGN.Basedonthesestudies,wemodeled
flatskylevel, increasingthe sizeofthe fitting regionwill the NC in NGC 4244 using two S´ersic components, and
lead to an increase in the brightness of the S´ersic com- for NGC 4395 we used a S´ersic component to model the
ponent,anincreaseinthe effective radius,anincreasein NC and a point source component to model the AGN.
the S´ersic index, and a decrease in the sky level. For ex- For comparison to the rest of the sample, we also per-
ample, when we performed a single S´ersic component fit formedsingleS´ersiccomponentfitstotheNGC4244and
8
F275W F336W F438W F547M F814W F127M F153M
FWHM=0.092" FWHM=0.093" FWHM=0.095" FWHM=0.097" FWHM=0.102" FWHM=0.191" FWHM=0.196"
Fig.4.—ModelPSFsineachfilter. ThefullwidthathalfmaximumofeachPSFisgiveninarcseconds.
NGC 4395 images. A more detailed discussion of indi- Because the data have very high S/N, the formal
vidual objects is given in Section 5. Figure 5 shows the GALFITmeasurementuncertaintiesinthebest-fitparam-
results of the fits to the F814W images for each NC. eters obtained from propagatingthe uncertainty in elec-
troncountsateachpixelarequite small. Ourerrorbud-
3.3. Comparison with ISHAPE get is dominated by uncertainties that arise due to the
degeneracybetweenthe NC andthe backgroundlight of
Georgiev & B¨oker(2014)usedtheISHAPEprocedurein
thehostgalaxylight,whichwemodelinGALFITusingei-
the BAOLAB software package (Larsen 1999) to measure
thera flatskylevelorthe combinationofa flatskylevel
effective radii, axis ratios, and magnitudes of a sample
and an exponential disk. Estimates of the uncertainties
of228NCsinlate-typespiralgalaxiesusingimagesfrom
of the best-fit parameters should therefore reflect how
the HST/WFPC2 archive. Much like GALFIT, ISHAPE
sensitive these parameters are to variations in the back-
measuresstructuralpropertiesbyminimizing the χ2 dif-
groundlevel. Ingeneral,thereisnounambiguouswayto
ference between an image and a PSF convolved model,
define a boundary between the NC and the underlying
butitperformsthefitoveracircular,ratherthanarect-
background light from the disk of the host galaxy. In
angularfittingregion. Intheirstudy,theNCsweremod-
addition, the background can be quite non-uniform and
eledusing single componentKing andpower-lawmodels
lumpy, due to the presence of resolved stars, off-nuclear
andtheHST/WFPC2modelPSFsweregeneratedusing
clusters,andpatchydustlanesinthe immediatevicinity
TinyTim. However,they didnot allowfor the possibility
of the NC.
of an exponential disk profile to model the host galaxy
Theformaluncertaintyonthe skyparameterreturned
light. Thesubstructureseeninsomeoftheclustersinour
by GALFIT is typically small (<0.1 ADU), and does not
samplewouldnotbeeasilydetectedatthedistancestypi-
reflectthe true uncertaintyresultingfromthe lumpy na-
calofNCsintheirsample,whichwasselectedusingadis-
ture of the immediate surroundings of the NC. In order
tancecutof40Mpc. Asatest,weusedISHAPEtofitsin-
to more realistically estimate the uncertainty in the sky
gleS´ersiccomponentmodelstoourF814Wimagesusing
parameter (σ ), we took the 401×401 pixel section of
a200pixelfittingradiusandfoundverygoodagreement sky
each image that was used in the fit, and divided it into
withourGALFITresults. Themeanoffsetbetweenmagni-
eightequally sizedwedges,by splitting it into quadrants
tudes measuredusing ISHAPEandmagnitudes measured
along the vertical and horizontal axes of the image, and
using GALFIT(i.e., |mGALFIT−mISHAPE|)was0.05mag, then splitting each quadrant in half at a 45◦ angle with
while on average, the effective radii and S´ersic indices
respect to the vertical and horizontal axes. We found
agreed within 11% and 24%, respectively.
the mean sky level in each wedge while masking the NC
and computed the standarddeviation of these values, to
3.4. 1D Radial Profiles
obtain σ . This uncertainty (typically a few ADU),
sky
The “goodness of fit” is easy to visualize in a 1D is much larger than the formal uncertainty on the sky
radial profile plot, although non-axisymmetric features parameter returned by GALFIT. We used the mean sky
of the surface brightness profiles are lost in this repre- level in this calculation, as opposed to the mode sky,
sentation. We fit elliptical isophotes (ellipses of con- because the mean sky is a better representation of the
stant surface brightness) to both the HST/WFC3 im- surface brightness of the host galaxy, which includes re-
ages and the PSF-convolved GALFIT models using the solved sources, such as bright stars and off-nuclear clus-
PyRAF/STSDAS routine Ellipse (Freudling 1993), ters. Using the mode sky would underestimate the host
which is based on methods detailed by Jedrzejewski galaxy surface brightness by excluding the contribution
(1987). For each isophote, this routine computes the of individual resolved stars.
semimajor axis (SMA) of the ellipse, as well as the For fits where a flat sky level was used to model the
isophote surface brightness and ellipticity (e = 1−b/a, background light from the host galaxy, the correspond-
where b/a is the axis ratio). We also performed isopho- ing uncertainties in the NC best-fit parameters were es-
talfitstothemodelPSFsinordertodisplaytheirradial timated by re-running GALFIT while holding the sky pa-
surface brightness profiles. Plots of the F814W radial rameterfixedatsky −σsky (wheresky isthebest-fitsky
profiles of the surface brightness and ellipticity for each parameter from the original fit), and again with the sky
cluster are shown in Figure 6. In every case the PSF- parameterfixedatsky +σsky. Forfitswhereacombina-
convolvedGALFITmodelsaresignificantlymoreextended tion of a flat sky leveland exponentialdisk were used to
thanthePSFandtheclustereffectiveradiusexceedsthe model the host galaxy, we re-ran the host galaxy fit as
radius which encloses half of the flux of the F814WPSF explainedinSection3.2,butheldtheskyparameterfixed
(0′.′063),indicating that allof the NCs in our sample are at sky − σsky, and again with the sky parameter fixed
spatially resolved. at sky + σsky. We then performed the NC fit with the
host galaxy exponential disk and sky parameters fixed
3.5. Uncertainties in Cluster Parameters at the best-fit values found in the previous step to find
9
IC 342
M33
NGC 247
NGC 300
NGC 2403
Fig. 5.—ResultsoftheF814Wfits. Thepanels(fromlefttoright)showtheF814Wimage,thebest-fitmodelconvolvedwiththePSF,
and the residuals (image−model). The field of view is 16′′×16′′ for all fits. Images are displayed with their native orientation, as in
Figure2.
10
NGC 2976
NGC 3621
NGC 4244
NGC 4395
NGC 7793
Fig.5.—ResultsofthefitsinF814W(continued).