Substellar Objects in Nearby Young Clusters (SONYC): The bottom of the Initial Mass Function in NGC1333
Alexander Scholz, Vincent Geers, Ray Jayawardhana, Laura Fissel, Eve Lee, David Lafreniere, Motohide Tamura
aa r X i v : . [ a s t r o - ph . S R ] J u l To appear in ApJ
Preprint typeset using L A TEX style emulateapj v. 03/07/07
SUBSTELLAR OBJECTS IN NEARBY YOUNG CLUSTERS (SONYC):THE BOTTOM OF THE INITIAL MASS FUNCTION IN NGC 1333 ∗ Alexander Scholz , Vincent Geers , Ray Jayawardhana , Laura Fissel , Eve Lee , David Lafreni`ere ,Motohide Tamura To appear in ApJ
ABSTRACTSONYC –
Substellar Objects in Nearby Young Clusters – is a survey program to investigate thefrequency and properties of substellar objects with masses down to a few times that of Jupiter innearby star-forming regions. Here we present the first results from SONYC observations of NGC 1333,a ∼ ∼ .
015 to 0.1 M ⊙ , based on model evolutionarytracks. For comparison, the completeness limit of our survey translates to mass limits of 0.004 M ⊙ for A V . M ⊙ for A V .
10 mag. Compared with other star-forming regions, NGC 1333shows an overabundance of brown dwarfs relative to low-mass stars, by a factor of 2-5. On the otherhand, NGC 1333 has a deficit of planetary-mass objects: Based on the surveys in σ Orionis, the OrionNebula Cluster and Chamaeleon I, the expected number of planetary-mass objects in NGC 1333 is8-10, but we find none. It is plausible that our survey has detected the minimum mass limit forstar formation in this particular cluster, at around 0.012-0.02 M ⊙ . If confirmed, our findings point tosignificant regional/environmental differences in the number of brown dwarfs and the minimum massof the Initial Mass Function. Subject headings: stars: circumstellar matter, formation, low-mass, brown dwarfs – planetary systems INTRODUCTION
The origin of the stellar Initial Mass Function (IMF)is one of the major issues in astrophysics. The low-massend of the IMF, in particular, has been the subject ofnumerous observational and theoretical studies over thepast decade (see Bonnell et al. 2007). While higher massstars exhibit the well-known Salpeter shaped mass func-tion ( dN ∝ m − α dm , with α = 2 .
35, Salpeter 1955), thepower-law slope becomes significantly shallower below ∼ . M ⊙ (Kroupa 2001; Chabrier 2003). In the sub-stellar regime, the coefficient α is mostly found to bearound 0.6 (e.g. Moraux et al. 2003; L´opez Mart´ı et al.2004; Moraux et al. 2007; Caballero et al. 2007). Whilethe higher-mass domain is thought to be mostlyformed through fragmentation (e.g. Padoan & Nordlund2002) and/or accretion onto the protostellar core(Bonnell & Bate 2006, e.g.), in the low-mass and substel- ∗ BASED ON DATA COLLECTED AT SUBARU TELESCOPE,WHICH IS OPERATED BY THE NATIONAL ASTRONOMI-CAL OBSERVATORY OF JAPAN.Electronic address: [email protected] SUPA, School of Physics & Astronomy, University of St. An-drews, North Haugh, St. Andrews, KY16 9SS, United Kingdom Department of Astronomy & Astrophysics, University ofToronto, 50 St. George Street, Toronto, ON M5S 3H4, Canada National Astronomical Observatory, Osawa 2-21-2, Mitaka,Tokyo 181, Japan ** Principal Investigator of SONYC lar regime additional physics is likely to play an impor-tant role, e.g. dynamical interactions (Bate et al. 2003;Bate 2009), turbulence (Padoan & Nordlund 2004),photoerosion in the radiation field of bright stars(Whitworth & Zinnecker 2004), or tidal shear in thegravitational potential of a stellar cluster (Bonnell et al.2008).To continue collapsing, an object or region has to coolas quickly as it is being heated by the conversion of grav-itational into thermal energy. As radiative cooling is re-lated to the opacity, this condition defines the opacitylimit of fragmentation, which translates into a minimummass for star-like sources, predicted to be between 0.001and 0.01 M ⊙ (e.g. Low & Lynden-Bell 1976; Rees 1976;Boss 2001; Bate 2005; Whitworth & Stamatellos 2006).Thus, isolated substellar objects in star-forming regionscan have planetary-like masses. Furthermore, currentnumerical simulations of star formation build a signifi-cant fraction of brown dwarfs (BDs) by fragmentationin protostellar disks and subsequent ejection from thedisk (Bate et al. 2003; Stamatellos & Whitworth 2009).Thus, a fraction of the substellar population may have asimilar origin to massive planets. Hence, using the term‘isolated planetary-mass object’ (IPMOs or planemos,see Basri & Brown 2006) for objects with masses belowthe Deuterium burning limit ( ∼ . M ⊙ ) makes somephysical sense, as this is the mass regime where star and Scholz et al.planet formation overlap.The frequency and physical properties of the ob-jects at the extreme low-mass end of the spectrum arepoorly known. Following the discovery of brown dwarfs(Rebolo et al. 1995; Nakajima et al. 1995), surveys ofstar-forming regions, young clusters, and the field haverevealed a rich population of substellar objects, trigger-ing a wealth of follow-up studies to characterize theirphysics (see Basri 2000). Populations of planetary-mass objects have been identified with masses downto 0 . − . M ⊙ (e.g. Zapatero Osorio et al. 2000;Lucas & Roche 2000). So far, there is no evidence fora cut-off in the mass function, or in other words, theopacity limit has not been observed yet.As of today, however, only two regions have been sys-tematically surveyed down to masses significantly belowthe Deuterium burning limit: σ Orionis cluster and theOrion Nebula Cluster (ONC). At a distance of ∼
400 pc,any follow-up for planemos in Orion beyond photom-etry and very low-resolution spectroscopy is challeng-ing. A few more objects in the same mass regime havebeen identified based on mid-infrared detections fromISO or Spitzer (Testi et al. 2002; Luhman et al. 2005;Allers et al. 2006; Luhman & Muench 2008). This ap-proach, however, will by definition only find sources withcircumstellar material and thus yield incomplete results.The unbiased census for the members of the ChamaeleonI (ChaI) star-forming region is currently complete downto ∼ . M ⊙ (Luhman 2007), and has yielded a handfulof planetary-mass members.SONYC – Substellar Objects in Nearby Young Clus-ters – is a new, unbiased attempt to establish the fre-quency of low-mass brown dwarfs and planetary-massobjects as a function of cluster environment and to pro-vide the groundwork for detailed characterization of theirphysical properties (disks, binarity, atmospheres, accre-tion, activity). The project is based on extremely deepoptical and near-infrared imaging and follow-up spec-troscopy using 8-m class telescopes, aiming to detect thephotospheres of the objects, which is a prerequisite foran unbiased survey. The observations are designed toreach limiting masses of 0.003 M ⊙ , and thus are meantto probe the opacity limit of star formation (see above).In each region, we aim to cover at least ∼ .In the first paper for this project, we report on thesurvey in the cluster NGC 1333, part of the Perseusstar-forming complex, carried out with Suprime-Camand MOIRCS at the Subaru telescope. With an ageof ∼ I = 25 mag and J = 21 mag, corresponding to masses for cluster mem-bers of about 0.003 M ⊙ , according to the COND03 evo-lutionary tracks (Baraffe et al. 2003). Furthermore, thecluster is compact and can thus be covered efficiently.Near-infrared surveys in NGC 1333 have been carriedout, among others, by Strom et al. (1976), Aspin et al.(1994), and Lada et al. (1996), establishing the presenceof an embedded stellar cluster and a site of ongoingstar formation with more than a hundred YSOs downto K = 14 . with a limitingmagnitude of K = 16 mag and follow-up spectroscopy. They find a disk frequency of 74% and an upper limitto the mass function exponent of α ≤ . ), probing for the first time the planetary-massregime. The near-infrared survey by Oasa et al. (2008)is similarly deep ( K = 18 mag) and covers 25 arcmin ,but does not include follow-up spectroscopy. In ad-dition, NGC 1333 has been covered by numerous X-ray, mid-infrared, sub-millimeter, and millimeter sur-veys, for example in the framework of the Spitzer ‘Coresto Disks’ Legacy Program, resulting in detailed stud-ies of the cloud and cluster structure as well as ac-cretion and outflow activity (e.g. Getman et al. 2002;Preibisch 2003; Enoch et al. 2006; Jørgensen et al. 2006;Walsh et al. 2007; Gutermuth et al. 2008).This paper is structured as follows: We present ourobservations and data reduction procedures for photom-etry and spectroscopy in Sect. 2. In Sect. 3 we discussthe photometric selection and spectroscopic verificationof very low-mass objects in NGC 1333. The properties ofour sources, including evidence for youth, are analysed inSect. 4. The low-mass end of the IMF based on our newfindings is discussed in Sect. 5. We give our conclusionsin Sect. 6. OBSERVATIONS AND DATA REDUCTION
Optical imaging and photometry
We observed NGC 1333 in the SDSS i’ and z’ filtersusing the Subaru Prime Focus Camera (Suprime-Cam)wide field imager (Miyazaki et al. 2002) on 2006 Novem-ber 17. The conditions were photometric with typicalseeing ranging from 0.45-0.63 ′′ . Suprime-Cam is a mo-saic camera which utilizes 10 CCDs arranged in a 5 by2 pattern giving a total field of view of 34 ′ × ′ . AsNGC 1333 is quite compact, the entire cluster was ob-served in a single pointing. A ten point dither patternwas carried out to eliminate gaps between the CCDs andto correct for bad pixels. We observed this pattern ineach filter six times, where the individual images haveexposure times of 60 sec, for a total integration time of3600 sec in each band.A standard reduction was performed for each individ-ual CCD chip. Routines from the Suprime-Cam datareduction software package SDFRED (Yagi et al. 2002;Ouchi et al. 2004) were used for overscan subtraction andflat fielding. In both filters, the images show significantextended structure from the molecular cloud emission.Sky subtraction was performed by fitting a Gaussian tothe peak of a histogram of pixel brightnesses and sub-tracting the centroid. Additional routines from
SDFRED were used for distortion correction, bad pixel maskingand image combination. Only one of the ten CCD chips( w67c1 ) showed noticeable fringing. To reduce each chipconsistently, and because the 10 point dither patternmostly removes the fringes, we did not carry out a fringecorrection. The world coordinate system was calibratedagainst sources from the 2MASS point source catalog(Cutri et al. 2003) with J ≥
13 using the msctpeak pro-gram from the IRAF package
MSCRED . The typicalfitting residuals were of order 0.25 ′′ . IRAF is distributed by the National Optical Astronomy Ob-
ONYC: NGC 1333 3
Fig. 1.— i’-band histogram of the objects in our photometriccatalogue with i’- and z’-band data. The peak at 24.7 mag indi-cates the completeness limit of the survey. The faintest objects inthe survey are found at i >
26 mag. The dashed and dash-dottedlines show the histogram for equally sized samples of blue and redobjects, illustrating that the completeness limit does not signifi-cantly change with color, although we are slightly more sensitiveto red objects.
Sources were identified from each CCD chip us-ing the Source Extractor (
SExtractor ) software package(Bertin & Arnouts 1996). For object extraction we re-quired at least 5 pixels to be above the 3 σ detectionlimit. SExtractor automatic aperture fitting photometryroutines were used to calculate the flux of each source.We rejected objects within a few pixels of the edge ofthe image, elongated objects ( a/b > .
2) and objectsthat were not present in both the i’ and z’ source lists.The rejection criteria were chosen conservatively, to min-imize the number of spurious detections in the photom-etry database. We visually checked the detection algo-rithm and confirmed that the number of point-sourcesthat are missed as well as the number of spurious sourcesthat are included is minimal. The final optical cataloguehas 7757 objects.We measured the chip-to-chip zeropoint offsets fromthe median fluxes of domeflat images. The absolute ze-ropoint for the mosaic was derived from observations ofthe SA95 standard field, which contains SDSS secondarystandards (Smith et al. 2002). To avoid saturation, theimages of standard fields were defocused. The fluxes offive standard stars were extracted using aperture pho-tometry with 50-pixel radius. We derived a mean ze-ropoint of 27 . ± .
01 in i’- and 27 . ± .
02 in z’-band. These values agree with the zeropoint measuredby Miyazaki et al. (2002) within 0 .
12 magnitude in i’ and0 .
07 in z’. The airmass difference between standard fieldsand science fields was ± .
1, i.e. the differential extinctionis negligible.From the histogram of the magnitudes (see Fig. 1for the i’-band data) we derived completeness limits ofour survey: In the i’-band, we are complete down to24 . ± . . ± . servatories, which are operated by the Association of Universitiesfor Research in Astronomy, Inc., under cooperative agreement withthe National Science Foundation. are close to the sensitivity limits determined duringthe Suprime-Cam commissioning phase (Miyazaki et al.2002). Near-infrared imaging and photometry
NGC 1333 was observed with the Multi-Object In-fraRed Camera and Spectrograph (MOIRCS) mountedon the Subaru Telescope, in the MKO broad-band fil-ters J and Ks, during the nights of 2008 December 4–6.MOIRCS uses two CCDs providing a total field of viewof 4 ′ × ′ with a spatial resolution of 0.117 arcsec/pixel.NGC 1333 was covered in 24 pointings in both filters,using a 6 or 8-point dither pattern. In the J-band, thetypical exposure time per pixel was 600 sec, in the K-band 300 sec. In addition, the center of NGC 1333 wascovered in a smaller series of 13 sec short exposures, toobtain photometry for the brightest sources that saturatein the deeper exposures.Data reduction was performed using a modified versionof the SIMPLE-MOIRCS package , which is written inIDL and uses SExtractor. Background subtraction, flat-field correction and distortion correction were performedon individual images, before the images were co-added.The world coordinate system was calibrated against the2MASS point-source catalogue. Initial source identifica-tion is performed using SExtractor, requiring at least 3pixels to be above the 3 σ detection limit. Source rejec-tion was performed based on saturation and elongation.After running the identification and rejection routines,the sources were visually verified for each field. Photom-etry was performed using a fixed-aperture extraction rou-tine in IDL. To account for variable seeing between fields,an aperture correction factor is determined for each in-dividual field. To correct for difference in sensitivity be-tween detector 1 and 2, a flux offset factor is determinedfrom the average level of the sky flux in both images, andapplied to detector 2 for each individual field.The final near-infrared catalogue has 2360 sources.Calibration of the absolute zeropoint magnitude is per-formed by comparing the full catalogue of extractedsources to the 2MASS point source catalogue. The meanzeropoint is derived to be 25.08 ± ± Multi-object spectroscopy
We used MOIRCS again to carry out multi-object spec-troscopy (MOS) for 53 sources in NGC 1333, 36 of themselected based on the photometry in the four bands i’, z’,J, K (see Sect. 3). The targets are covered in six multi-slit masks; the number of objects per mask was 5-13.Pre-imaging for the masks in the K-band was obtainedin September 2008. We used slits which are 0.9 ′′ wideand 9-12 ′′ long, except for a few longer slits, used to ob-serve standard stars. The spectroscopy run was carriedout in three nights on Dec 4-6, 2008. For each maskwe integrated in total 50-60 min with the grism HK500,split in shorter exposures of 5 or 10 min, depending onconditions. Between the single exposures, the mask was ∼ whwang/idl/SIMPLE/MOIRCS/Doc/ Interactive Data Language
Scholz et al.moved so that the targets shifted by 2.5 ′′ along the slit(nodding), to facilitate sky subtraction. The seeing dur-ing the science integrations was stable at 0.5-0.7 ′′ . Be-fore and after the NGC 1333 masks, we observed A0stars through one of the science masks for flux calibra-tion and extinction correction; these exposures requiredde-focusing to prevent saturation. For each mask, weobtained series of domeflats with lamp on and off forcalibration purposes.The MOS data was reduced following standard recipesfor near-infrared spectra. In a first step, we subtractedthe pairs of nodded exposures, which removes the skybackground and the detector bias. These images weredivided by a normalized flatfield, which is the differencebetween the averaged lamp-on and lamp-off domeflats.Frames from the same nodding position were co-added.This gives us two final images for each chip and mask.In these frames, the signal-to-noise ratio for the spectraranges from 20 for the faintest to 1000 for the brightestsources. These values have been measured in the centralparts of the H- and K-band, in regions without strongskyline residuals.For the extraction of the spectra we used the apall rou-tine within IRAF. All objects from our photometric can-didate list are well-detected. For wavelength calibrationwe additionally extracted the spectra from the unreducedframes, which contain plenty of telluric OH lines. As thewavelength solution depends slightly on the position ofthe slit on the chip, the fit was done for each object sep-arately. For the HK grism, a second order polynomial fitcovering about 20 skylines gave a typical rms of 2-3 ˚A,well below the resolution. All spectra are smoothed witha ±
25 ˚A boxcar average and binned to 40 ˚A per pixel.The two nodded spectra for the same object were coad-ded. The standard star spectra were treated in exactlythe same way as the science frames.Near-infrared spectra show strong effects of the atmo-sphere – broad absorption features, mostly between theJ/H and H/K bands, as well as sharp OH emission lines.In our data, subtracting the nodding exposures reducedthe strength of the OH lines by 95% or better. The resid-uals are removed in the background subtraction whichis part of the extraction. The corrections for telluricextinction and instrument response are usually done inone step, based on observations of a standard star withknown spectral energy distribution. Our A0 standardstar spectra were divided by a library spectrum of anA0 star (Pickles 1998), obtained from the ESO website .To remove residuals of the Hydrogen absorption featuresseen in A0 spectra, we linearly interpolated from 1.54to 1.78 µm and from 2.09 to 2.34 µm , regions that aremostly free from additional telluric absorption bands.All science spectra were divided by one of these modi-fied standard spectra observed at similar airmass. Typ-ically, the airmass difference is < .
1. Still, we find thatthe correction of atmospheric correction is not optimum.We attribute this mainly to the duration of our scienceexposures: While the standards were observed within afew minutes, the science exposures cover 1-2 hours. Vari-able sky conditions, as they have been present during theobservations, hamper a good extinction correction. For our analysis we therefore focus on the spectral regionsleast affected by telluric absorption, i.e. 1.5-1.75 µm inthe H-band and 2.1-2.3 µm in the K-band. SELECTION OF NEW SUBSTELLAR MEMBERS INNGC 1333
Optical color-magnitude diagram
The goal of the project is to identify new very low-mass members of NGC 1333. Due to their pre-mainsequence status, plus extinction, these sources are ex-pected to occupy a distinct area in optical/near-infraredcolor-magnitude diagrams, on the red side of the broadcumulation of background main sequence stars. In Fig.2 we show the (i’, i’–z’) diagram, which we used toidentify the initial list of candidates. The cumulationof non-members around i ′ − z ′ = 1 . i ′ ∼ . i ′ ∼ . i ′ . . Near-infrared color-magnitude diagrams
We used the near-infrared photometry to verify theinitial candidate selection. In particular, we aim to en-ONYC: NGC 1333 5
Fig. 2.—
Color-magnitude diagram in (i’, i’–z’) constructed from our deep Suprime-Cam observations. The selection box for follow-up spectroscopy is shown with dash-dotted lines. Objects chosen for spectroscopy (plus) and confirmed or possible M-type membersof NGC 1333 (cross, see Table 1 and 2) are marked. Spectroscopically confirmed BDs from the surveys by Wilking et al. (2004) andGreissl et al. (2007) are marked with diamond and triangle symbols respectively. sure that our candidates show the colors expected forreddened very low-mass members of NGC 1333. In Fig.3 we show the color-magnitude diagram constructed bycombining the Suprime-Cam i’-band photometry withthe MOIRCS J-band data. The diagram shows all ob-jects with data in both bands, but objects selected forspectroscopy and the 28 objects confirmed by the spec-troscopy (Sect. 3.3) are marked. A few of the confirmedobjects are not in our near-infrared catalogue due to sat-uration; for them we use 2MASS data instead.For comparison, we overplot the BCAH98, COND03,and DUSTY00 evolutionary tracks for 1 and 10 Myrfrom Baraffe et al. (1998); Chabrier et al. (2000);Baraffe et al. (2003) and the photometry for the sub-stellar population in the σ Ori cluster (Caballero et al.2007). In both cases, we converted the Cousins I-banddata to the Sloan i’-band using the transformationgiven by Jordi et al. (2006). In principle this requiresknowledge of the R − I color, which is not alwaysavailable for the σ Ori sources. We assumed an average R − I = 2 . σ Ori objects may therefore be incorrect by about ± .
2. Inconsistencies between the Suprime-Cam i’-bandwith the standard bandpass may cause additionaluncertainties. The σ Ori photometry was shifted to thedistance of NGC 1333, assuming a differential distancemodulus of 0.7, consistent with recent estimates. σ Oriis a cluster with little extinction; we expect these objectsto be mostly free from reddening. The plot demonstrates that the spectroscopy candi-dates cover the color space that is limited on the blueside by models and the σ Ori sources. This is as ex-pected, since the cloud cores in NGC 1333 have signifi-cant and variable extinction so the population of youngcluster members as well as background objects should bescattered towards the red side. Apart from reddening,there is no systematic offset between our sample and thecomparison data. Thus, we confirm that our selectionof candidates covers the color space expected for verylow-mass members of NGC 1333.The intrinsic near-infrared color J − K of late-typeobjects is mostly independent of spectral type and lu-minosity; thus it can be used to estimate the extinc-tion for the suspected cluster members. This is doneonly for the objects confirmed as having spectral typeM or later (Sect. 3.3). We assume here an intrinsiccolor of J − K = 1 . ± . M ⊙ ), BCAH98 (for 0.03-0.5 M ⊙ ),and DUSTY00 (for 0.006-0.1 M ⊙ ), and with the colors ofbrown dwarfs in the 3 Myr old σ Ori star-forming regiondown to 0.013 M ⊙ (Caballero et al. 2007). For youngobjects with masses below the Deuterium burning limitthere are some indications that the J − K color increasesto 1-1.5 (Lodieu et al. 2006; Caballero et al. 2007); thismay lead to an overestimate of the extinction by as much Scholz et al. Fig. 3.—
Color-magnitude diagram in (J, i’–J) constructed from our MOIRCS and Suprime-Cam observations. The candidates selectedfor spectroscopy from the optical diagram and the confirmed M-type objects (see Table 1 and 2) are marked with plusses and crosses,respectively. For comparison, we overplot model isochrones for 1 and 10 Myr and the photometry for the substellar objects in σ Ori(Caballero et al. 2007), shifted to the distance of NGC 1333. For the models and the σ Ori objects the Cousins-I photometry was shiftedto the Sloan-i’ band using the transformation given by Jordi et al. (2006).
Fig. 4.—
Near-infrared color-magnitude diagram for the 28 ob-jects listed in Table 1 and 2. The plusses show the MOIRCS pho-tometry, crosses are 2MASS values for the cases not covered by ourdata, the circles the dereddened values, assuming intrinsic colorsof J − K = 1 . as 2.5 mag for the lowest mass objects.For the extinction estimate and throughout this paper,we use the extinction law from Mathis (1990), with R V =4. In Fig. 4 we plot the ( J , J − K ) color-magnitudediagram with the solid line indicating the reddening pathand the dotted line showing the assumed intrinsic color.For the 28 objects with spectroscopic confirmation (listed in Tables 1 and 2), we plot our photometry (crosses) or2MASS data (plusses), as well as the dereddened J-bandphotometry (circles). For the optical extinction in ourconfirmed sample, we obtain A V = 0 −
13, i.e. we areprobing deeper into the cloud than Wilking et al. (2004, A V ≤ A V are given in Tables 1 and2. This range of optical extinctions is in good agreementwith the extinction maps as derived from submillime-ter continuum observations (Enoch et al. 2006). Con-verting a 1.1 mm map to an extinction map results in A V <
17 mag for 5 ′ resolution for NGC 1333. Our ex-tinction estimates for individual objects are all in thisrange, perhaps with the exception of object CO observations, mainly becausethese indicators do not trace the densest regions of thecloud well.The main error sources in the extinction estimates area) uncertainties in the estimate of the intrinsic colors andb) possible excess emission due to disks and accretion.As pointed out above, except for the lowest mass ob-jects, the intrinsic colors are likely to be accurate within0.2 mag, which causes errors in A V of < A V is 2 mag; larger errors are possible inindividual cases. The extinction values will be used inthe spectral fitting in Sect. 3.3. Spectroscopy
With the multi-object spectroscopy described in Sect.2.3 we obtained low-resolution spectra for 53 objects intotal. The sample of objects with spectra includes 36candidates from the primary photometric sample dis-cussed in Sect. 3.1. In addition, we have spectra fornine objects that are slightly bluer in optical colors thanour primary candidates sources, to verify the validity ofthe optical color criterion (called ‘blue’ objects in the fol-lowing). Among our spectroscopy sample are nine con-firmed young M dwarfs from the sample published byWilking et al. (2004), six of them also in our own photo-metric candidate sample. Finally, we included five ran-dom objects in our MOS fields. We had 12 additionalslits on objects that are not well detected and did notallow a reliable extraction. All these undetected sourceshave been included as either blue or random objects, anddo not come from our photometric sample.Our goal is to identify young members of NGC 1333with substellar masses, i.e. with spectral types later thanM5. In addition to these young brown dwarfs, our sam-ple may contain embedded stellar members of NGC 1333and reddened background stars with spectral types ear-lier than M5, and (less likely) late M and early L typefield objects in the foreground. We show some examplesof the expected shape of the spectra for reddened youngbrown dwarfs, derived from model spectra, in Fig. 5.The low resolution of the spectra does not allow us tomeasure narrow atomic features. Instead, we focus on thebroadband appearance, i.e. the shape of the H- and K-band as well as the relative intensity in these two bands.We benefit from the fact that the overall shape of thenear-infrared spectrum changes distinctly at early/midM spectral types. While earlier type stars show a broad-band spectrum which is monotonously decreasing to-wards longer wavelength, this is not the case for > M5objects, due to the appearance of strong H O absorptionfeatures at 1.3-1.5 µm and 1.75-2.05 µm . Most notably,this causes the H-band spectrum to have a distinct andbroad peak centered on 1.68 µm , in contrast to earliertype objects, which increases in strength towards L-typeobjects. This fundamental change is discussed in detailin the literature, see for example Cushing et al. (2005).As demonstrated in Fig. 5, the H-band peak is dominantfor temperatures lower than 3000 K. While the strengthand shape of the H-band peak is primarily determined bytemperature (and gravity, see 4.2), there is an additionaleffect of extinction – high extinction (see Sect. 3.2), willmake the H-band peak slightly sharper. As can be seen inFig. 5, this is clearly a minor effect and is only apparentwhen comparing over a wide range of extinctions.All spectra were examined visually to look for the typ- Fig. 5.—
Reddened model spectra from the AMES DUSTY se-ries (Allard et al. 2001) for log g = 3 .
5, corresponding to pre-mainsequence ages. Shown are three different temperatures (from top tobottom: 3000 K, 2400 K, 1800 K), each with three different extinc-tions A V = 0 , ,
10 mag as solid, dotted, dashed lines. The H-bandpeak and its dependence on temperature is clearly visible. The rel-ative strength of H- and K-band is strongly affected by extinction.Dash-dotted lines show the spectral range in our observations usedfor classification. ical signature of water absorption, mainly the character-istic peak in the H-band. The sample clearly falls intotwo groups, where 28/53 objects (53%) show an indi-cation of a peak while the others do not. In this clas-sification, we erred on the side of caution, and there-fore also included objects with only tentative evidenceof water absorption. This sample of 28 objects is the‘confirmed’ sample of Figs. 2-4 and is listed in Tables 1and 2, together with the photometric and spectroscopicproperties. All objects without a clear H-band peak aremost likely stars with spectral types significantly earlierthan M5. Our spectra do not permit definitive classi-fication of these sources. Out of the 28 confirmed ob-jects, 21 come from our photometric candidate sample.All nine objects from Wilking et al. (2004) with spectraltypes M2-M8 are correctly classified as having an H-bandpeak. The ‘peak’ objects exhibit other spectral featuresof late-type sources, most notably the CO bandheads at > . µm . We do not find evidence for CH absorptionat 1.6-1.8 µm , typical for late L and T dwarfs.We compared the 28 spectra with the H-band ‘peak’with the latest version of the spectra from the AMESDUSTY model atmospheres (Allard et al. 2001), which Scholz et al. Fig. 6.—
Spectra of late-type objects in NGC 1333 (solid lines) in comparison with models (dashed lines), first part. The runningnumber is the same as in Tables 1 and 2. Shown is the best fit model with the effective temperature as given in the tables. Objects 1-19are probable substellar members of NGC 1333. are based on the PHOENIX stellar atmosphere code. Incontrast to the NEXTGEN spectra from the same code,DUSTY spectra include dust formation in the convectiveenvelope and dust opacities. The more recent DUSTYmodels also include more complete H O and TiO linelists. To probe if the specific treatment of dust is rele-vant, we compared the DUSTY spectra with the DRIFTPHOENIX series (Schmidt et al. 2008), which include adust model and consistently treat cloud formation andits feedback on the atmosphere. For the resolution andwavelength coverage comparable with our observed spec-tra, there is no significant difference in the temperatureregime > g (3.5 or 4.0). The model spectra were binned andsmoothed in the same way as the science spectra. Allspectra are scaled to a flux of 1.0 at 1.7 µm . Each ob-ject spectrum was then compared with the model series,reddened with the optical extinction derived in Sect. 3.2.The quality of the match was judged by calculating the χ aided by visual inspection. To avoid being misled byexcess continuum due to disks or accretion and uncer-tainties in the extinction, we put emphasis on matchingthe H-band shape, rather than the overall spectrum. In Figs. 6 and 7 we show all 28 object spectra withH-band peak (solid lines) and their best fit model spec-trum (dashed lines). The ID numbers are the same as inTable 1 and 2. The best match temperatures are givenin the tables. For 18 out of 28 objects, this procedureyields a minimum χ of < . A V ≤ A V = 4 . T = 3800 K. As argued in Sect. 4.1, the status of this ob-ject is unclear; it may be a red giant in the backgroundof the cluster. Object A V = 18 mag.The source is only partially covered in the H-band, there-fore the temperature estimate is preliminary. Based onthe slope of the H-band spectrum, the best estimate is T = 2200 − Fig. 7.—
Spectra of late-type objects in NGC 1333 (solid lines) in comparison with models (dashed lines), second part. The runningnumber is the same as in Tables 1 and 2. Shown is the best fit model with the effective temperature as given in the tables. Objects 1-19are probable substellar members of NGC 1333. between the model and observed spectra. In Fig. 8 thetypical uncertainty of the fitting procedure is illustratedby comparing the observed spectra for three objects fromTable 1 with models for various effective temperatures.The χ values used to select the best-fitting model arelisted as well. As can be seen in the figure, the procedureoften does not allow us to distinguish reliably betweentwo models with ∆ T = 100 K. However, for ∆ T &
200 Kthere are discernible differences, particularly in the H-band peak, between model and observation, combinedwith a clear increase in χ . Thus, a conservative estimatefor the typical uncertainty in T eff is ±
200 K. For all > M5objects, the spectral types – either from Wilking et al.(2004) or from the spectral index, see below – are con-sistent with the temperatures from model fitting within ±
200 K or ± > µm , as expected for a mid/late M-type object.Thus, we consider the temperature estimate of ∼ µm , consistent with the 1-0 S(1) line ofmolecular hydrogen at 2.122 µm , seen in protostellar out-flows. At λ > . µm , the flux decreases sharply, whichis not reproduced by any models or literature spectra inthis temperature domain. Based on the H-band peak,our best estimate for the effective temperature is 2600 K,but this has to be treated with caution.As a complementary approach, we computed the spec-tral index proposed by Allers et al. (2007) for spectralclassification of young BDs, applicable from M5 to L5.This index has two distinct advantages compared withthe variety of other index schemes in the literature, asit is gravity insensitive and does not require wavelengthcoverage in regions heavily affected by telluric water ab-sorption. The spectral types from this index confirm theWilking et al. (2004) spectral types for the seven > M5objects within ± . Summary of substellar member selection
Fig. 8.—
Illustration of the spectral fitting process: Each panel shows the spectrum of a probable brown dwarf in NGC 1333 fromTable 1 in comparison with model spectra for a range of effective temperatures. For clarity, the models are offset by 0.05 units. For eachcomparison, we give the effective temperature of the model and the χ of the fit. From the χ values we infer a conservative uncertaintyof ±
200 K for the effective temperature. The plot demonstrates that larger offsets in T eff lead to visible discrepancies between model andobservation, particularly in the shape of the H-band peak. Note: For object A V = 2 . χ ≥ .
37; much better results are achieved with a slightly lower A V . We use here A V = 1 mag. In summary, from our 36 photometrically selected can-didates, we confirm 21 as M-type objects. In addition,three out of nine bluer objects and one out of five ran-dom objects are classified as M-type sources. Nine clus-ter members from Wilking et al. (2004) are confirmedby our spectroscopy, seven of them overlapping with ourphotometrically selected sample. In total, we have 28confirmed objects with a H-band peak, indicative of spec-tral type M. In this confirmed sample are 19 objects withspectral types > M5 or effective temperatures of ≤ CHARACTERIZATION OF CONFIRMED LATE-TYPEOBJECTS
Effective temperatures
We have determined effective temperatures for thesample of late-type objects in Tables 1 and 2 by fit-ting the observed to model spectra. Out of 28 sources,19 have temperatures between 2500 and 3000 K, for theother nine we obtain higher values. In addition, we havealready derived extinctions, based on the J − K colors,and dereddened J-band magnitudes (see Sect. 3.2, Fig.4).We use the dereddened J-band photometry to obtainan alternative estimate of effective temperatures by com-paring with the BCAH98 evolutionary tracks BCAH98(Baraffe et al. 1998) for the typical cluster age of 1 Myr.This relies on the assumption that the objects are in fact young members of the cluster NGC 1333, for whichwe adopt a distance of 300 pc, consistent with the es-timates by de Zeeuw et al. (1999); Belikov et al. (2002).The J-band is the wavelength regime that is normallyassumed to be least affected by accretion, disk excess,and magnetic activity. It is found that the BCAH98temperature vs. J-mag relation is well approximated bya fourth degree polynomial. Applying this function tothe dereddened J-band photometry yields ‘photometric’temperatures from 2100 to 3400 K.For 15 objects, the photometric and spectroscopic tem-peratures agree within ±
200 K, confirming our assess-ment of T eff given in Sect. 3.3. Four objects (23, 26-28in Table 2) have photometric temperatures 800-1200 Klower than the ones from the spectroscopy, which can beexplained with being background objects. Assuming in-trinsic colors for dwarf stars from the 1 Gyr isochrone inthe BCAH98 model tracks puts their distances at 360-950 pc. The remaining objects have offsets of 200-500 Kbetween the two values, and in most cases the photomet-ric temperatures are too low. One possible explanation isan age spread in the cluster. As shown by Winston et al.(2009), the members of NGC 1333 are significantly dis-persed in the HR-diagram, possible indicating an agespread from < TABLE 1Probable substellar members of NGC 1333. no. α (J2000) δ (J2000) i’ (mag) z’ (mag) J (mag) K (mag) A V T eff a SpT b Notes c d,e d e d,e
11 03 29 07.17 +31 23 22.9 22.989 21.187 17.87 15.62 6.6 (2600) MBO141 e
12 03 29 09.33 +31 21 04.1 22.553 20.336 16.33 13.29 10.7 (2500) MBO70, ex e
13 03 29 10.79 +31 22 30.1 19.483 17.778 14.96 13.05 4.8 3000 M7.5 MBO6214 03 29 14.43 +31 22 36.2 18.104 16.960 14.55 13.02 2.8 2900 M7 MBO6615 03 29 17.75 +31 19 48.1 19.091 17.326 14.80 12.99 4.3 3000 M7.5 MBO64, ex d,e
16 03 29 28.16 +31 16 28.4 16.934 15.808 13.22 12.53 0.0 2600 M7.5 random17 03 29 33.88 +31 20 36.1 20.383 18.848 16.47 15.37 0.5 2500 M8 MBO14018 03 29 35.71 +31 21 08.5 22.633 21.089 18.50 16.94 3.0 2500 blue, ex19 03 29 36.36 +31 17 49.8 22.528 20.716 17.91 16.38 2.8 2700 a Estimated by comparing the spectra to models. Uncertain fit results are in brackets. b Estimated using the spectral index and calibration from Allers et al. (2007), if applicable. c ‘ex’ means mid-infrared color excess, ASR and MBO object id’s are candidates from Wilking et al. (2004). d Near-infrared photometry from 2MASS. e Temperature estimate changed after visual inspection, see Sect. 3.3.
TABLE 2Possible stellar members of NGC 1333. no. α (J2000) δ (J2000) i’ (mag) z’ (mag) J (mag) K (mag) A V T effa SpT b Notes c
20 03 28 43.23 +31 10 42.6 21.864 19.606 15.37 12.03 12.3 3400 ex d,e
21 03 28 47.34 +31 11 29.9 21.626 19.330 15.48 12.70 9.4 3100 d
22 03 28 50.97 +31 23 47.9 20.380 18.418 14.66 11.24 12.7 3500 MBO26 d
23 03 29 02.16 +31 16 11.1 17.197 16.245 14.53 13.57 0.0 3900 ASR03, giant? d
24 03 29 08.32 +31 20 20.3 19.376 17.809 14.80 12.48 7.0 3100 MBO44 e
25 03 29 11.64 +31 20 37.5 20.544 18.738 15.43 12.67 9.3 3400 MBO58 d
26 03 29 16.75 +31 23 25.3 15.42 13.58 4.4 3800 MBO79, ex, giant? d,e
27 03 29 22.36 +31 21 36.8 21.407 20.197 17.71 15.89 4.3 3300 MBO177, blue28 03 29 33.50 +31 17 56.5 22.926 21.477 18.52 16.27 6.6 3700 blue a Estimated by comparing the spectra to models. Uncertain fit results are in brackets. b Estimated using the spectral index and calibration from Allers et al. (2007), if applicable. c ‘ex’ means mid-infrared color excess, ASR and MBO object id’s are candidates from Wilking et al. (2004). d Near-infrared photometry from 2MASS. e Temperature estimate changed after visual inspection, see Sect. 3.3. than for dwarfs (about M0 or earlier for these two ob-jects). These types would be consistent with our effec-tive temperatures. Thus, these two objects could be redgiants in the background of NGC 1333. However, assum-ing the intrinsic giant colors from Bessell & Brett (1988),this would put them as an unreasonably large distanceof 30 kpc for spectral type M0. A significantly earlierspectral type is at odds with our spectroscopy. Thus,the nature of these objects remains uncertain.
Cluster membership
For the likely brown dwarfs in our sample (see Table1), the best evidence for youth and thus cluster mem-bership is given by the shape of the H-band peak inthe spectra. For substellar objects, the H-band peakappears to be gravity sensitive. Low gravity and thusyouth is signified by a distinctly sharp peak, as op-posed to the more rounded peaks seen in old objects (Kirkpatrick et al. 2006). The same characteristic is alsoseen in the available model spectra (Allard et al. 2001).All 19 sources with effective temperatures of 3000 Kor lower listed in Table 1 show a sharp peak. As anexample, we show four of them in comparison with pub-lished spectra for field dwarfs of similar spectral typefrom the NIRSPEC Brown Dwarf Spectroscopic Survey(McLean et al. 2003) in Fig. 9, left panel. The differencein the shape of the H-band feature is obvious. Particu-larly, the slope on the red side of the peak is significantlysteeper in our objects. On the other hand, our spectraagree well with those of young brown dwarfs in Taurus,taken from Muench et al. (2007), as shown in Fig. 9,right panel.Statistically, we do not expect a significant contami-nation with late field M dwarfs in our sample in Table1. According to the data compiled in Caballero et al.(2008), the space density of M6-7 dwarfs is 4 . · − pc − . · − pc − for M8-9. Our samplecovers an area of about 200 arcmin which correspondsto ∼ . at the distance of NGC 1333. In a cone withthis footprint and 300 pc height we expect about one M6-M9 field dwarf. Thus, for the M6 or later objects in Table1 the contamination by field objects is negligible.We conclude that the > M5 objects identified in thissurvey (‘probable’ members in Table 1) are in fact youngsubstellar members of the cluster NGC 1333. For theearlier type stars (‘possible members’ in Table 2), sig-nificant contamination by field dwarfs or giants is morelikely, as their space density is larger and the H-bandsignature of youth is less pronounced. Further evidencefor youth comes from the fact that many of our objectsshow mid-infrared color excess indicative of the presenceof circumstellar material, as discussed in Sect. 4.5, andfrom the spatial clustering around the main cloud corein NGC 1333, see Sect. 4.4.
Masses
We estimate the masses of the probable brown dwarfsin our sample by comparing the effective temperaturesand the dereddened J-band photometry with the theo-retical model tracks for an assumed age. Considering theuncertainties in model tracks and object ages, we refrainfrom deriving masses for individual objects. Instead, wefocus on establishing the approximate mass limits in oursample.The range of effective temperatures in this sample is2500-3000 K. Comparing directly to the 1-3 Myr tracksfrom BCAH98 yields masses of 0.02-0.1 M ⊙ . The resultis very similar for the COND03 and DUSTY00 tracks. Ifsome objects are younger than 1 Myr, they will be slightlymore massive for a given temperature. Taking into ac-count 200 K uncertainty in our temperature estimate, thelowest mass limit becomes 0.012-0.015 M ⊙ , according tothe COND03 and DUSTY00 tracks. Based on the dered-dened J-band photometry, the mass range in our browndwarf sample is 0.006-0.1 M ⊙ for the 1 Myr isochrones, or0.012-0.2 M ⊙ for the 5 Myr isochrones. Note that the lat-ter values are not significantly affected by uncertainties inthe distance to NGC 1333. The most recent distance esti-mates agree fairly well at ∼
300 pc with an uncertainty ofabout 30 pc (de Zeeuw et al. 1999; Belikov et al. 2002),corresponding to ± . M ⊙ based on its J-band magni-tude (assuming a distance of 435 ±
55 pc); the actualsystem mass is 0.09 M ⊙ . More indirectly, Mohanty et al.(2004) find that the models predict brown dwarfs to betoo bright in J-band at a given mass, by up to one or-der of magnitude. Both tests indicate that the massesfrom the model tracks, when inferred from luminosities,may systematically underestimate the actual masses bya factor of up to two. Based on the comparison of effective temperatures withthe three model tracks, we conclude that the lowest masslimit in our brown dwarf sample is 0.012-0.02 M ⊙ . At thehigh mass end, the mass range of the objects in Table 1extends across the substellar boundary into the stellarregime. Thus, a few of these objects may turn out to bevery low-mass stars, instead of brown dwarfs. Spatial distribution and binarity
In Fig. 10 we show the position of our photometric can-didates and the confirmed members of NGC 1333 in rela-tion to the gas in the cloud and other samples of clustermembers. As already pointed out in Sect. 3.1, the photo-metric candidates show a strong clustering towards thecenter of the cloud. On one side, this can be explainedwith the increasing extinction, causing background ob-jects to appear heavily reddened. On the other hand, asseen in the plot, YSOs candidates selected from Spitzermid-infrared excess (Gutermuth et al. 2008) show thesame strong clustering, indicating that the distributionof members is very compact and does not extend to theedges of our survey field. Our spatial coverage thus fullyencompasses the extent of the cluster. The small num-ber of photometric candidates in the outer region is alsoevidence for the lack of contamination in our sample, asargued in Sect. 4.2.All our spectroscopy fields were located within ± . ′′ of each of ourspectroscopically confirmed M-type sources. The lowerseparation limit for the detection of companions is setby the seeing and the rejection criterion for elongatedobjects (see Sect. 2.1); we determine this to be at ∼ ′′ (Sect. 2.1), corresponding to a separation range of300 to 1000 AU for the distance of NGC 1333. No com-panions were found. For substellar objects in NGC 1333(Table 1) this gives a 2 σ upper limit for binarity in thegiven separation range of 19%.A number of recent studies clearly show a dearth ofONYC: NGC 1333 13 Fig. 9.—
Spectra of very low-mass objects in NGC 1333 from Table 1 (solid) in comparison with spectra of field dwarfs from the NIRSPECBrown Dwarf Spectroscopic Survey (left panel, McLean et al. 2003) and with young brown dwarfs in Taurus (right panel, Muench et al.2007). The field dwarfs are from bottom to top Wolf 359 (M6), VB 10 (M8), LHS 2065 (M9); the Taurus BDs are from bottom to topMHO 4 (M7), KPNO 1 (M7.5), and KPNO 5 (M8.5). The objects in NGC 1333 have a significantly ‘sharper’ peak than the field dwarfs, butare well matched by the young Taurus BDs. To account for the higher extinctions in NGC 1333, the literature spectra have been reddenedto A V = 5 mag. wide binaries in the very low-mass regime. In the centralparts of the Cha I star forming region, for example, abinary fraction of 11% is found for separations >
20 AU(Ahmic et al. 2007), and zero systems with separationslarger than 50 AU. Our result in NGC 1333 is consistentwith these findings (for a review on VLM binarity seeBurgasser et al. 2007).
Spitzer photometry
The NGC 1333 cluster is part of the Perseus star-forming complex and thus has been covered by theSpitzer Legacy Program ‘From Cores to Disks’ (C2D, PI:Neil Evans). From the C2D source catalogues (versionS13, full Perseus catalogue) we obtained the IRAC andMIPS fluxes from 3.6 to 24 µm for all objects in Tables1 and 2. Out of 28 sources, 16 are well-detected with er-rorbars < . µm , with magnitudes > µm . This confirms youth and thus mem-bership to NGC 1333 for a large fraction of our sample.Objects with mid-infrared color excess are marked with ‘ex’ in Table 1. The disk fraction in our (small) sam-ple is ∼ THE MASS FUNCTION IN NGC 1333
Ratio of stellar to substellar members
Combining our survey with the ones by Wilking et al.(2004) and Greissl et al. (2007) there is now a substan-tial sample of spectroscopically confirmed brown dwarfsin NGC 1333. We use this to improve the constraintson the ratio of low-mass stars to brown dwarfs, a num-ber recently measured for a range of other star-formingregions. We identify 19 objects as young M6 or latersources. Seven of them have previously been confirmedby Wilking et al. (2004). Their survey has found ninemore objects with M6 or later spectral type, which we4 Scholz et al.
Fig. 10.—
Spatial coverage of our survey and distribution of objects. Crosses are all candidates selected from optical photometry with i ′ > . CO from the J = 1 − Fig. 11.—
IRAC color-color diagram for the 16 objects fromTables 1 and 2 which have errorbars < . consider to be good candidates for brown dwarfs. Thestudy by Greissl et al. (2007) has limited spatial cover-age (3 arcmin ), but finds five more objects in this spec-tral regime. All these objects are located close to thecluster center, which points to a high likelihood of beingmembers. Thus, the current total census of likely browndwarfs in this cluster is 33. As there are a number ofadditional photometric candidates in the cluster region (see Fig. 10) that lack spectroscopic confirmation, thisnumber should be treated as a lower limit; on the otherhand, as pointed out in Sect. 4.3, some of these objectsmay be slightly above the substellar boundary.To constrain the number of stellar members inNGC 1333, we use the Spitzer selected sample byGutermuth et al. (2008), which identifies 137 likely YSOsin the cluster, 93 of them also have 2MASS near-infraredmagnitudes. From these, 91 are in our survey area. Wedereddened the photometry for these objects based onthe J − K color, using the procedure discussed in Sect.3.2. Comparing the dereddened J-band magnitude withthe BCAH98 evolutionary track allows us to select ob-jects in a given mass range. In this sample, 38 sourcesare likely to have masses between 0.08 and 1 M ⊙ . Basedon the spectroscopic follow-up by Winston et al. (2009),this sample is unlikely to have significant field star con-tamination. Spitzer only finds objects with disks; in fact,the cluster is known to harbour a significant populationof disk-less T Tauri stars (Winston et al. 2009). Assum-ing a disk fraction of 83% (Gutermuth et al. 2008) givesa total number of ∼
50 stellar cluster members in thismass regime. Taking into account the uncertainties inour mass cuts, this number may be inaccurate by up to20%.Although the Spitzer sample of stars has been identi-fied in a way not comparable to our own brown dwarfsurvey (and the two other ones in the literature), we areconfident that it gives a robust estimate for the numberONYC: NGC 1333 15of stars in this mass range in the cluster. The range ofextinctions and the spatial distribution are similar in theSpitzer sample and the brown dwarf sample. Moreover,the disk fraction among the brown dwarfs is similar tothe one seen in stars (Sect. 4.5). Thus, in terms of evo-lutionary state and spatial coverage the stellar sample isrepresentative.From these numbers, we estimate the ratio of (low-mass) stars (0 . < M < . M ⊙ ) to brown dwarfs(0 . < M < . M ⊙ ) in NGC 1333 to be 50 / . ± .
3. This can directly be compared withthe literature. In an IMF analysis for seven differ-ent star-forming regions (not including NGC 1333),Andersen et al. (2008) find this ratio to be 3.3-8.5 de-pending on the region, with the lowest values measured inthe ONC, NGC 2024, and Chamaeleon. Taking into ac-count the 1 σ uncertainty in their values, the minimum ra-tio becomes 1.9 in Chamaeleon. These numbers are stillsignificantly higher than the ratio in NGC 1333, pointingto an overabundance of brown dwarfs in NGC 1333 by afactor of 2-5.This result is consistent with the previous findings forthis cluster. Wilking et al. (2004) derived a ratio ofstars (0 . < M < . M ⊙ ) to very low-mass objects(0 . < M < . M ⊙ ) of 1 . ± . . , and considered this tobe a lower limit, i.e. fully comparable with our result.(This would become 2.0, if they assume an age of 1 Myrinstead of 0.5 Myr.) Including their lower mass browndwarfs, Greissl et al. (2007) find a value of 0 . ± . The minimum mass in NGC 1333
The depth and large spatial coverage of our surveyallows us to put meaningful limits on the abundanceof planetary-mass objects in NGC 1333 in comparisonwith other regions. Recapitulating, the coolest objectsin NGC 1333 have effective temperatures of 2500 K, cor-responding to masses of 0.012-0.02 M ⊙ . That means thatthere is no planetary-mass object in our spectroscopicallyconfirmed sample. In the following, we will compare withthe mass completeness limits in our survey and the fre-quency of planemos in other star-forming regions. Giventhe considerable uncertainties in the model evolutionarytracks in this age and mass regime, we will follow twoguidelines: a) Conversion from effective temperatures tomasses is more reliable than from luminosities (or mag-nitudes) to masses, see the discussion in Sect. 4.3. b)We prefer to base the arguments on relative comparisonsconsistently using the same model tracks, and do notplace strong emphasis on the absolute mass values.Our coolest and lowest mass objects in the spectro-scopic sample have T eff = 2500 ±
200 K, i.e. we ruleout the presence of objects with 2200 K or cooler inthis group of objects. The five faintest confirmed browndwarfs have 22.5-23.5 mag in the i’-band with extinctions A V .
10 mag. Our completeness limit is 24.7 mag ini’-band, i.e. 1.2-2.2 mag deeper than the faintest con-firmed objects. Based on the 1 Myr isochrones fromCOND03 and DUSTY00, an object that much fainter than one with 2500 K has 0.006-0.01 M ⊙ , with T eff of2000-2200 K, i.e. for A V .
10 mag we are completedown to ∼ . M ⊙ . For A V . ∼ . M ⊙ or 1800 K. Thus, our survey is com-plete to masses significantly below the low-mass limitof the confirmed objects. In fact, six objects with i’-mag > . σ Ori cluster, the Orion Nebula Cluster (ONC), andthe Chamaeleon I star-forming region (Cha I). The mostrecent census of the substellar population in σ Ori, pub-lished by Caballero et al. (2007), lists 34 objects with0 . < M < . M ⊙ and 12 with lower masses. Intheir mass scale, the planetary-mass objects have spec-tral types M9 or later, indicating that they exhibit lowertemperatures (and thus lower masses) than the coolestobjects in NGC 1333. The numbers of BDs and planemosin σ Ori are robust against the particular choice of evo-lutionary tracks and uncertainties in age, distance, andreddening. According to Caballero et al. (2007), the cen-sus is expected to contain ∼ ∼ σ Ori is8/32 or 25 ± %. This is a lower limit, as the surveydoes not show any cut-off of the mass function downto 0 . M ⊙ , their completeness limit. According to re-cent adaptive optics observations in the core of σ Ori(Bouy et al. 2009) the survey by Caballero et al. (2007)might miss a number of objects with planetary massesclose to the bright central star.In the ONC, the highly variable extinction makes anunbiased census difficult. In the largest spectroscopicsurvey in the substellar regime to date, Weights et al.(2009) have confirmed ∼
38 objects with masses at or be-low the Hydrogen burning limit. According to their massestimates, 17 objects in this sample have 0 . < M < . M ⊙ and 20 objects have even lower masses. Themasses are based on a comparison of the observed HR di-agram with evolutionary tracks (including COND03 andBCAH98, as used in this paper) and thus rely partly onspectroscopic temperatures, comparable to our estimatesin NGC 1333. Deriving the mass limits from the dered-dened H-band photometry, as given by Weights et al.(2009), yields 15 objects with 0 . < M < . M ⊙ and14 below (assuming a distance of 450 pc). Their analy-sis of the mass function thus supports the presence of aplanetary-mass population in the ONC.There are two alternative estimates for the frequencyof planemos in the ONC in the literature: According toLucas et al. (2006), 7.5% (1-14%) of the total populationin the cluster have a mass of 0.003-0.015 M ⊙ , where thewide range of possible values is mostly a reflection forage uncertainty. Lucas et al. (2005) find a drop by factortwo in the mass function at the Deuterium burning limit,a deviation from a flat mass function in the substellarregime. Their upper limit for the fraction of planemos inthe total cluster population is 10-13%.For Cha I, the most up-to-date unbiased census byLuhman (2007) yields a total cluster population of 226objects, among them 28 brown dwarfs, and four sourceswith masses below the Deuterium burning limit. Based6 Scholz et al.on these numbers, the planemo fraction is 2% relative tothe total population and the planemo vs. brown dwarfratio is 17%. In this region, however, the census is com-plete only down to ∼ . M ⊙ . The mass function isflat in the substellar regime and does not show indica-tions for having reached the minimum mass (Luhman2007). Thus, the total number of planemos may be twiceas high, depending on the cut-off of the IMF. In fact,Luhman & Muench (2008) published one more sourcein the planetary-mass regime based on a Spitzer mid-infrared detection.Based on these literature findings, we can estimatethe expected number of planemos in NGC 1333. Thisis illustrated in Fig. 12. In our survey area, we find19 brown dwarfs, seven of them previously identified byWilking et al. (2004). In addition, the area covers 14more substellar objects from other sources (see Sect. 5.1).With one exception, all these brown dwarfs exhibit opti-cal extinctions <
10 mag. For this extinction value, oursurvey is complete down to ∼ . M ⊙ . The numberof confirmed planetary-mass objects in the survey lim-its is zero. Based on the total sample covered in ourspectroscopic observations, the 1 σ upper limit for thisnon-detection is about 2 objects.Assuming the planemo vs. brown dwarf ratio as seen inthe σ Ori cluster, we would expect to find 8 ± σ Ori. The ONC surveyby Weights et al. (2009) predicts a number of planemosroughly comparable with the number of brown dwarfs,i.e. ∼
33 for NGC 1333. Based on the two other esti-mates, and assuming a total cluster population of ∼ <
20, most likely ∼
10 planemos in the cluster. Basedon the mass function in Cha I, we expect 6-12 planemosin NGC 1333. Thus, all estimates for the expected num-ber of planetary-mass objects are as high as or higherthan the 2 σ limit derived from our survey in NGC 1333.We conclude that NGC 1333 likely shows a deficit ofplanemos compared with all other star-forming regionswhere comparable survey depth has been reached.We arrive at essentially the same conclusion when weuse temperatures instead of masses: The coolest objectin our survey has an effective temperature of 2500 K,corresponding to a spectral type of M9 (Luhman et al.2003). In σ Ori, there are 12 objects known withlater spectral types (Barrado y Navascu´es et al. 2001;Mart´ın et al. 2001); a significant fraction of the low-mass population is not spectroscopically confirmed. TheONC survey by Weights et al. (2009) gives 17 sourceswith > M9 and effective temperatures < ∼
300 more area coverage, encompassing the full ex-tent of the cluster. This result is even more remarkablewhen we take into account that the number of browndwarfs in NGC 1333 is unusually high, as pointed out inSect. 5.2.The best explanation for this outcome is a significantdrop in the mass function in NGC 1333 around the Deu-terium burning limit, corresponding to effective tempera-tures below 2500 K in this cluster. Compared with otherclusters like σ Ori and the ONC, the mass function ofNGC 1333 does not extend down to planetary masses.As our survey is deeper than previous work in other re-gions and covers a substantial sample of objects withfollow-up spectroscopy, this is the best evidence so farfor a cut-off in the mass function at very low masses. Itis plausible that our survey has detected the minimummass limit for star formation in this particular cluster,at around 0.012-0.02 M ⊙ .With the currently available sample, this is a 2 σ re-sult and thus needs further verification. If confirmed, itwould point to strong environmental dependences in themass function, particularly the mininum mass of star for-mation. In all other regions surveyed with comparabledepth, the mass function extends well below 0.01 M ⊙ ,while the spectrum is truncated at ∼ . − . M ⊙ inour target cluster NGC 1333.At this point it is not clear what could cause a changein the minimum mass by a factor of two or more. Com-pared with regions as diverse as σ Ori, the ONC, andChaI, there is no evidence that NGC 1333 is differ-ent in any significant way in terms of initial conditions,dust properties, or metallicity. Based on the currentlyavailable census, the object density in NGC 1333 isslightly higher than in Cha I and significantly higherthan in σ Ori. While the number of low-mass mem-bers in all three regions is probably comparable andin the range of 150-250 objects, the radii of their dis-tribution are 0.3 pc in NGC 1333, 0.5-0.7 pc in Cha I,and 3-5 pc in σ Ori (Sherry et al. 2004; Luhman 2007;Gutermuth et al. 2008). Bonnell et al. (2008) recentlysuggested that brown dwarfs in clusters form through thegravitational fragmentation of infalling gas. This couldexplain the overabundance of substellar objects in thedenser NGC 1333, but would also predict a higher num-ber of planetary-mass objects. On the other hand, thedifferences in density in those three regions are, as ex-plained in Sect. 4.5, most likely a consequence of pro-gressing dynamical evolution and thus a difference in age,and not necessarily in initial conditions. Thus, a consis-tent interpretation of our findings in terms of the forma-tion process is currently lacking. CONCLUSIONS
We present the first results of the SONYC project,short for ‘Substellar Objects in Nearby Young Clusters’.We have obtained deep optical and near-infrared obser-vations of the NGC 1333 cluster using the Subaru 8-mtelescope. From the photometry in the i’, z’, J and Kband, we have selected brown dwarf candidates. Follow-up infrared spectroscopy, also obtained at Subaru, is usedto verify their nature. In the following, we summarize themost important findings of this survey: • In our large-scale, optical+near-infrared imagingsurvey, we reach completeness limits of 24.7 in i’-ONYC: NGC 1333 17
Fig. 12.—
Mass distribution of brown dwarfs in NGC 1333 (hatched histogram) and the deficit of planetary-mass objects. The threelabelled datapoints show the predicted number of planemos in NGC 1333 based on the surveys in σ Ori, the ONC, and in ChaI. Errorbarscorrespond to 1 σ . The x-values for these three datapoints roughly correspond to the completeness limits in the literature surveys for theseregions. The non-detection of planetary-mass objects points to a deficit of objects in this mass regime compared with other clusters. Thedotted lines refer to our detection limit for Av = 5 and 10 mag.. band, 20.8 in J-band, and 18.0 in K-band. In termsof object masses for members of NGC 1333, thiscorresponds roughly to mass limits of 0.008 M ⊙ for A V .
10 mag and 0.004 M ⊙ for A V . • From the optical photometry, we select 196 ob-jects with i’–z’ colors in the range as expected forsubstellar and planetary-mass cluster members, in-cluding a number of previously identified browndwarfs. 36 of these objects are chosen for follow-upspectroscopy. This spectroscopic sample does notshow any bias in spatial coverage or optical/near-infrared colors with respect to the full photometriccandidate sample. • Using multi-object spectroscopy in H- and K-band,we confirm 28 objects close to the cluster center aslate-type sources, 21 of them from the sample se-lected from photometry. This is based on water ab-sorption features in the H-band. The extinctions,derived from the J − K color, for the 28 objectscover the range from 0 to 13 mag. Fitting modelspectra to the observed ones, we find effective tem-peratures of 2500-3900 K with a typical uncertaintyof ±
200 K. • From the 28 confirmed late-type objects, 19 haveeffective temperatures of 3000 K or lower (and spec-tral types M6 or later). Combined with the clearindications for youth, this classifies them as proba-ble substellar members of NGC 1333. Of this sub-sample, seven have been previously identified byWilking et al. (2004). The remaining nine objectsin the confirmed sample are possible steller mem-bers, which likely exhibit significant contaminationby field dwarfs/giants. • The confirmed objects are strongly clusteredaround the core of NGC 1333. The majorityof them shows evidence for disks in Spitzer mid-infrared colors. Both findings additionally provideevidence for membership. • Combining our survey results with previous stud-ies, the current census of spectroscopically con-firmed brown dwarfs in NGC 1333 is 33. For com-parison, the number of stellar members with massesbetween 0.08 and 1 M ⊙ is in the range of 50. Thisindicates a clear overabundance of brown dwarfs inNGC 1333; the ratio of substellar to stellar mem-bers with masses below 1 M ⊙ is 1 . ± .
3, lower bya factor of 2-5 than in all other previously surveyedregions. • The low-mass limit of the confirmed brown dwarfsin NGC 1333 is 0.012-0.02 M ⊙ , but the complete-ness limits are at significantly lower masses. Thus,8 Scholz et al.although being able to do so, we do not find apopulation of planetary-mass objects in this clus-ter. Scaling from literature results in other regions( σ Ori, ONC, Cha I), we would expect to find 8-10 ‘planemos’, but we find none. This indicatesa cut-off in the mass spectrum around the Deu-terium burning limit in NGC 1333; we have possi-bly reached the minimum mass of the IMF. • If confirmed, these last two findings point to re-gional/environmental variations as a major factorin determining the number of very low-mass ob-jects and the minimum mass of the IMF. It isparticularly difficult to explain simultaneously theoverabundance of brown dwarfs and the deficit ofplanetary-mass objects in NGC 1333. Thus, our re-sults present a challenge for the current theoreticalframework of star formation. • The mass spectrum in the brown dwarf and plane-tary mass regime necessarily relies on evolutionarymodels. Although we have made every effort toderive robust conclusions, we cannot get aroundthe fact that evolutionary models lack reliable cal-ibrations at the lowest masses. This is particularly true at young ages, where initial conditions mayaffect mass estimates, for example. Thus, the lackof objects with
T <
REFERENCESAhmic, M., Jayawardhana, R., Brandeker, A., Scholz, A., vanKerkwijk, M. H., Delgado-Donate, E., & Froebrich, D. 2007,ApJ, 671, 2074Allard, F., Hauschildt, P. H., Alexander, D. R., Tamanai, A., &Schweitzer, A. 2001, ApJ, 556, 357Allen, L. E., Calvet, N., D’Alessio, P., Merin, B., Hartmann, L.,Megeath, S. T., Gutermuth, R. A., Muzerolle, J., Pipher, J. L.,Myers, P. C., & Fazio, G. G. 2004, ApJS, 154, 363Allers, K. N., Jaffe, D. T., Luhman, K. L., Liu, M. C., Wilson, J. C.,Skrutskie, M. F., Nelson, M., Peterson, D. E., Smith, J. D., &Cushing, M. C. 2007, ApJ, 657, 511Allers, K. N., Kessler-Silacci, J. E., Cieza, L. A., & Jaffe, D. T.2006, ApJ, 644, 364Andersen, M., Meyer, M. R., Greissl, J., & Aversa, A. 2008, ApJ,683, L183Aspin, C., Sandell, G., & Russell, A. P. G. 1994, A&AS, 106, 165Baraffe, I., Chabrier, G., Allard, F., & Hauschildt, P. H. 1998,A&A, 337, 403—. 2002, A&A, 382, 563Baraffe, I., Chabrier, G., Barman, T. S., Allard, F., & Hauschildt,P. H. 2003, A&A, 402, 701Barrado y Navascu´es, D., Zapatero Osorio, M. R., B´ejar, V. J. S.,Rebolo, R., Mart´ın, E. L., Mundt, R., & Bailer-Jones, C. A. L.2001, A&A, 377, L9Basri, G. 2000, ARA&A, 38, 485Basri, G., & Brown, M. E. 2006, Annual Review of Earth andPlanetary Sciences, 34, 193Bate, M. R. 2005, MNRAS, 363, 363—. 2009, MNRAS, 392, 590Bate, M. R., Bonnell, I. A., & Bromm, V. 2003, MNRAS, 339, 577Belikov, A. N., Kharchenko, N. V., Piskunov, A. E., Schilbach, E.,& Scholz, R.-D. 2002, A&A, 387, 117Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393Bessell, M. S., & Brett, J. M. 1988, PASP, 100, 1134Bonnell, I. A., & Bate, M. R. 2006, MNRAS, 370, 488Bonnell, I. A., Clark, P., & Bate, M. R. 2008, MNRAS, 389, 1556Bonnell, I. A., Larson, R. B., & Zinnecker, H. 2007, in Protostarsand Planets V, ed. B. Reipurth, D. Jewitt, & K. Keil, 149–164Boss, A. P. 2001, ApJ, 551, L167Bouy, H., Hu´elamo, N., Mart´ın, E. L., Marchis, F., Barrado YNavascu´es, D., Kolb, J., Marchetti, E., Petr-Gotzens, M. G.,Sterzik, M., Ivanov, V. D., K¨ohler, R., & N¨urnberger, D. 2009,A&A, 493, 931 Burgasser, A. J., Reid, I. N., Siegler, N., Close, L., Allen, P.,Lowrance, P., & Gizis, J. 2007, in Protostars and Planets V,ed. B. Reipurth, D. Jewitt, & K. Keil, 427–441Caballero, J. A., B´ejar, V. J. S., Rebolo, R., Eisl¨offel, J., ZapateroOsorio, M. R., Mundt, R., Barrado Y Navascu´es, D., Bihain, G.,Bailer-Jones, C. A. L., Forveille, T., & Mart´ın, E. L. 2007, A&A,470, 903Caballero, J. A., Burgasser, A. J., & Klement, R. 2008, A&A, 488,181Chabrier, G. 2003, PASP, 115, 763Chabrier, G., Baraffe, I., Allard, F., & Hauschildt, P. 2000, ApJ,542, 464Cushing, M. C., Rayner, J. T., & Vacca, W. D. 2005, ApJ, 623,1115Cutri, R. M., Skrutskie, M. F., van Dyk, S., Beichman, C. A.,Carpenter, J. M., Chester, T., Cambresy, L., Evans, T.,Fowler, J., Gizis, J., Howard, E., Huchra, J., Jarrett, T.,Kopan, E. L., Kirkpatrick, J. D., Light, R. M., Marsh, K. A.,McCallon, H., Schneider, S., Stiening, R., Sykes, M., Weinberg,M., Wheaton, W. A., Wheelock, S., & Zacarias, N. 2003,2MASS All Sky Catalog of point sources. (The IRSA 2MASSAll-Sky Point Source Catalog, NASA/IPAC Infrared ScienceArchive. http://irsa.ipac.caltech.edu/applications/Gator/)de Zeeuw, P. T., Hoogerwerf, R., de Bruijne, J. H. J., Brown,A. G. A., & Blaauw, A. 1999, AJ, 117, 354Enoch, M. L., Young, K. E., Glenn, J., Evans, II, N. J., Golwala,S., Sargent, A. I., Harvey, P., Aguirre, J., Goldin, A., Haig, D.,Huard, T. L., Lange, A., Laurent, G., Maloney, P., Mauskopf,P., Rossinot, P., & Sayers, J. 2006, ApJ, 638, 293Folha, D. F. M., & Emerson, J. P. 1999, A&A, 352, 517Getman, K. V., Feigelson, E. D., Townsley, L., Bally, J., Lada,C. J., & Reipurth, B. 2002, ApJ, 575, 354Greissl, J., Meyer, M. R., Wilking, B. A., Fanetti, T., Schneider,G., Greene, T. P., & Young, E. 2007, AJ, 133, 1321Gutermuth, R. A., Myers, P. C., Megeath, S. T., Allen, L. E.,Pipher, J. L., Muzerolle, J., Porras, A., Winston, E., & Fazio, G.2008, ApJ, 674, 336Helling, C., Dehn, M., Woitke, P., & Hauschildt, P. H. 2008, ApJ,675, L105Jordi, K., Grebel, E. K., & Ammon, K. 2006, A&A, 460, 339