A population of very young brown dwarfs and free-floating planets in Orion

Abstract

1 Introduction

The last five years have seen great progress in the detection of brown dwarfs in the local neighbourhood, young Galactic clusters and star formation regions, starting with the near-simultaneous discovery of the first clearly confirmed brown dwarfs, Teide 1 in the Pleiades (Rebolo, Zapatero-Osorio & Martin 1995) and Gl229b in the local neighbourhood (Nakajima et al. 1995). Star formation regions offer the advantage that substellar objects are three orders of magnitude more luminous at an age of a few Myr than at an age of a few Gyr. Early photometric and spectroscopic work (Comeron et al. 1993; Williams et al. 1995; Comoron, Rieke & Rieke 1996) indicated that brown dwarfs are probably very common in star formation regions. Deep imaging of the Chamaeleon I dark cloud (Tamura et al. 1998; Oasa, Tamura & Sugitani 1999) has led to the discovery of two free-floating objects which may well have masses similar to those of planets. However, confirmation of substellar status is problematic in star formation regions, owing to the ubiquity of lithium in young objects and the complicating effects of extinction on both photometry and spectroscopy.

Recently, high-quality spectroscopy (Luhman & Rieke 1998; Luhman et al. 1998; Wilking, Greene & Meyer 1999) and the publication of theoretical evolutionary models for young substellar objects (Burrows et al. 1997, hereafter B97; D'Antona & Mazzitelli 1997, updated electronically 1988, hereafter DM98) have provided convincing evidence that photometric identification of young brown dwarf candidates is reliable. In photometric studies, masses of candidate objects are derived by comparison of the observables (luminosity and temperature) with the evolutionary tracks. The isochrones of B97 and DM98 are in fairly good agreement with regard to the mass-luminosity relation at an age of about 1 Myr, but there is some disagreement about the mass-temperature relation (Hertzsprung-Russell diagrams are compared by Luhman & Rieke 1998). Even if the theoretical effective temperatures were without flaw, there would be considerable uncertainty in the derivation of temperatures from photometry or spectroscopy, at the level of ±200 K in M and L dwarfs. Hence we use luminosity, which is more easily measured, to derive masses for our sources.

In this paper we report the results of a deep infrared photometric survey of the Trapezium cluster in Orion. A large population of substellar objects is discovered, including a number of free-floating objects of planetary mass. We note that the group at the Instituto de Astrofísica de Canarias (IAC; Bejar et al. in preparation) has simultaneously reported a similar discovery of planetary-mass objects in the adjacent σ Orionis cluster. The Trapezium has been intensively studied for many years and we have been able to draw upon a large body of publications to aid in our work. We selected the Trapezium for several reasons. (1) Its very high stellar density allowed photometry of several hundred sources in a fairly small survey. (2) False positive detections are essentially eliminated because the dense backdrop of OMC-1 obscures all background stars even in the K band, as shown by Hillenbrand & Hartmann (1998) through optical-infrared comparison of the cluster stellar density profile. (3) The extinction within the cluster is relatively low (0<A_V<15 for most sources), permitting reasonably precise dereddening. (4) Star formation is essentially complete in the cluster and the age range is thought to be 0.3−2 Myr, so the age-luminosity degeneracy is not large.

2 Observations

Deep Imaging of the Trapezium cluster was carried out at the United Kingdom Infrared Telescope (UKIRT) on 1998 December 14–16 and 22–23, using UFTI, the UKIRT Fast Track Imager. The observations of December 22–23 were made by observatory staff to compensate for time lost to poor weather and equipment failures on December 15–16. UFTI is a high-resolution camera constructed by the authors at the University of Oxford with assistance from several other UK institutions (Roche & Lucas 1998). It has a 1024×1024 HAWAII array sensitive between 0.78 and 2.5 μm. The image scale is 0.091 arcsec pixel⁻¹, yielding a field of view of 92.6 arcsec². Observations of 15 contiguous fields were made in the I, J and H filters, with 900-s exposures in each filter. Twilight flat-fields were taken to reduce the data. Seeing conditions were typically 0.6 arcsec FWHM in all three filters, with only slightly poorer image quality in the I band. The fine pixel scale led to a high degree of oversampling, which was very useful for distinguishing stars from small knots of nebulosity and in permitting reliable photometry of low signal-to-noise ratio detections.

2.1 Filter selection

Taking advantage of the short-wavelength sensitivity of the HAWAII array, UFTI contains I_U- (0.786–0.929 μm) and Z_U-band (0.85–1.05 μm) filters. These filters sample the steeply rising part of brown dwarf spectra from the optical to the near-infrared flux peak. The I_U filter was selected because the Z_U band contains a very strong [S iii] emission line which might have contaminated the photospheric flux in ‘proplyd’ sources, and because the (I–J) colour is a more reliable temperature indicator than the smaller (Z−J) colour given that fluxes have to be dereddened. The J and H filters were chosen to measure extinction using the temperature-insensitive (J−H) colour. The K filter was not selected because a significant fraction of the flux comes from hot circumstellar dust at 2.2 μm, making it unreliable for determining extinction or luminosity, and because it contains fairly strong low-excitation emission lines.

The UFTI I_U-band calibration (equation 1 below) relative to Cousins I was made by observation of eight faint red standards of approximately solar metallicity taken from Leggett, Allard & Hauschildt (1998). The calibration has been confirmed by observatory staff, including observations of blue standards, and can be fitted remarkably well by a linear equation. The I_U bandpass is somewhat redder than I_C, which compensates for the low quantum efficiency of the HAWAII array at these wavelengths (∼23 per cent) when observing cool stars. The J and H filters are of the new type commissioned by the Mauna Kea Consortium. The transformation to the CIT colour system [equations (2a) and (b) below] was determined by combining the transformations of S. Leggett for CIT to UKIRT_IRCAM with that for UKIRT_IRCAM to UKIRT_UFTI. We use the UFTI J and H magnitudes in this paper, but use Cousins I for ease of comparison with other studies. This is possible because fortuitously the net effect of extinction on equation (1) is much less than the measurement errors.

(1)

(2a)

(2b)

2.2 Data reduction and photometry

The data were reduced using iraf and the Starlink package ccdpack for image mosaicking. Photometry was carried out using the daophot package in iraf. Photometry in the core of the Trapezium cluster is complicated by the pervasive nebulosity, which has structure on all the observed spatial scales. This often leads to inaccurate measurement of sky background when doing automated photometry, and causes the daofind algorithm to misidentify many small-scale nebular flux variations as stars. To overcome these problems, we cross-correlated the stars found in each filter to remove most of the nebulous sources, and also rejected all very blue sources [(J−H)<0.2], which inspection indicated were all spurious. Every source was then visually inspected and photometry was performed manually in order to select the best appropriate sky annulus in each case. The results of manual photometry were generally used in preference to the results of automated crowded-field photometry, with a few exceptions where the allstar routine was needed for photometry of close binaries. The final photometric precision is ∼5 per cent in the outlying regions of the survey where the nebulosity is faint, limited by temporal variations in the image profile. Precision in the bright nebulous core (in which the majority of sources lie) is between 5 and 20 per cent (for the worst cases), depending on the degree of spatial variation of surface brightness and the source magnitude. Since a large fraction of young stars are variable, more precise photometry at one epoch would not have had much greater value.

3 Results

The main results of the survey are presented graphically in Fig. 1 (left- and right-hand panels). 515 unsaturated point sources were detected in both the J and H bands, of which 313 were also detected in the I band. An additional 48 sources were detected at greater than 5σ in the H band alone, and appear as upper limits in the right-hand panel of Fig. 1. An approximate colour-magnitude sequence for zero extinction is indicated by the dotted line. The value of the nearly temperature-independent intrinsic (J−H) colour over the range ∼2200–4000 K is clear, since nearly all low-mass stars and substellar sources will deredden to a colour near (J−H)=0.6. However, we prefer to use a two-colour sequence for formal dereddening where possible (see Section 3.2). The empty region to the upper right of the diagram represents the saturation limit near m_H=11.7.

Figure 1.

Left-hand panel: J versus (J−H) plot. Open circles are highly uncertain data points. The dotted line is an approximate zero-reddening track (see text). The solid lines are parallel to the A(V)=7 reddening vector and divide the population into stars, brown dwarfs and planets, using the B97 prediction and an age of 1 Myr. The dashed lines correspond to the 0.3- and 2-Myr predictions, indicating the effect of the age spread on the classification. The effect is similar at the planetary boundary. Right-hand panel: H versus (J−H) plot. This includes upper limits for sources with J>20, which confirm the paucity of faint blue sources.

Open in new tabDownload slide

A large fraction (≈32 per cent) of the JH sources are brown dwarf candidates, lying below the reddening track for a 0.08-M_⊙ star at an age of 1 Myr, as calculated from the B97 isochrones. Approximately 13 sources appear to lie below the 1-Myr track for an object with the minimum mass to burn deuterium (≈0.013 M_⊙). Following the definition suggested by B97, it is convenient to call such objects free-floating planets, even though they are likely to have formed by cloud core fragmentation in the same manner as stars and brown dwarfs. A definition by mass has the advantage that it can be applied before all the formation mechanisms are known. An interesting feature of the left-hand panel of Fig. 1 is the paucity of very faint blue sources with m_J>19, given that several fairly red sources are seen below this limit. This observation is based on a small number of objects, but it is supported by the upper limits in the right-hand panel, which show that many very faint red sources are detected at H band only, but none with (J−H)<1.3. This appears to indicate a sharp drop in the cluster luminosity function, at a level corresponding to about 8M_Jup (Jupiter masses). The significance of this is discussed in Section 4.2.

IJH photometry of the observed point sources is available electronically (by FTP to star.herts.ac.uk, in directory pub/Lucas/Orion), together with astrometry, dereddened magnitudes, luminosities, derived masses (using both the B97 and DM98 isochrones) and temperatures. We have surveyed the inner regions of the Orion nebula cluster, in which most star formation is thought to have occurred over the period between 0.3 and 2 Myr ago (Ali & Depoy 1995; Hillenbrand 1997), which leads to a small uncertainty in the derived masses which we have indicated in the left-hand panel of Fig. 1 by plotting the brown dwarf threshold for three different ages. A small proportion of younger sources is undoubtedly present, the masses of which will be less than we have estimated, but we have excluded the 18 sources that exhibited extended (non-stellar) profiles from our photometry and performed a further colour selection against proplyds (see Section 3.2), so only a few extremely young sources should remain in our IMF analysis.

The I versus (I–J) data in Fig. 2 show that nearly 90 per cent of the sources follow a well-defined sequence which is almost parallel to the reddening line. However, 41/313 of the sources (13 per cent) have much bluer (I–J) colours and lie to the left of the arbitrary line, parallel to the reddening line, which is plotted in Fig. 2. An initial suspicion that this might be due to poor photometry was disproved by observing that the same two sequences are followed in a subset of 92 sources located more than ∼2 arcmin away from the cluster core, where photometry is not compromised by bright nebulosity (not shown). Since we have selected filters with no strong low-excitation emission lines, the obvious interpretation is that the blue colours are due to scattered light from those objects that are very young or are viewed close to the plane of the accretion disc. This interpretation is confirmed by comparison with the list of proplyds detected via emission-line imaging with the Hubble Space Telescope (HST) in O'Dell & Wong (1996). Only 11/41 were detected in their relatively shallow optical surveys, which suffer from greater extinction and do not overlap precisely with our survey. However, 10/11 are listed as proplyds, and 1/11 is listed as a star. The star is presumed to illuminate circumstellar matter that was not detected by HST because it does not receive sufficient UV radiation from the central O-type stars, or because the proplyd ‘tail’ lies too close to the line of sight. These 10 proplyds are marked with crosses in Fig. 2.

Figure 2.

I versus (I–J) plot. Anomalously blue sources lie to the left of the arbitrary line parallel to the reddening track. Of the 11 blue sources detected by HST, 10 are proplyds, plotted as crosses.

Open in new tabDownload slide

The 41 blue sources are widely distributed throughout the survey region and do not display obviously unusual (J−H) colours, only a weak tendency to be bluer than the rest. The I versus (I–J) diagram appears to be an efficient way of detecting proplyds at large distances from the photoionizing O stars. However, the effect of circumstellar matter on infrared colours is not obvious, since it depends on the orientation of the system and distribution of matter in a complicated way (e.g. Kenyon et al. 1993). The observed (J−H) colours may well be different from the photospheric colours in these systems, so we have excluded all the blue sources from our dereddening analysis. It is likely that some sources in the red group also have slightly modified colours owing to anomalous extinction (i.e. both absorption and scattering), but this is not expected to be significant for most sources of age ∼1 Myr, since the spectral energy distributions of T Tauri stars are usually well fitted by a Planckian (photospheric) function at wavelengths between the visible and the thermal infrared (e.g. Rydgren, Strom & Strom 1976; Wilking, Lada & Young 1989). The 202 faint sources detected only at J and H cannot be probed for anomalous colours in this way and probably include some sources that are therefore inaccurately dereddened. However, the proportion of anomalous sources will be lower than 13 per cent in this group because the blue sources are easier to detect at I band. Those sources in the left-hand panel of Fig. 1 with m_J>16.2 that are detected in all three filters are almost exclusively members of the blue group. Hence blue sources detected only at J and H will be very few except below the I-band detection limit which corresponds to m_J≈18. A likely example of one such faint source at lower left in Fig. 1 is Orion 131-047 [adopting the O'Dell & Wen (1994) naming convention], for which m_J=18.22 and (J−H)=0.39. This is an unrealistically blue colour for such a low-luminosity source (which has accurately measured fluxes), so it is likely that the (J−H) colour has been reduced by at least 0.2 mag by scattering effects.

3.1 Extinction law

Many observers have found evidence for an anomalous extinction law in the Trapezium at optical and infrared wavelengths (e.g. Davis, Larson & Hofmann 1986; Cardelli, Clayton & Mathis 1989, hereafter CCM). We adopt the R_V=5.5 extinction law of CCM, calculated for θ₁C Ori using data at similar wavelengths to those observed here. The infrared extinctions for this law are A(I_C)=0.643, A(I_U)=0.583, A(J)=0.334 and A(H)=0.214 for A(V)=1. The effect of the unusually high R_V is to increase the slope of the reddening lines, resulting in higher derived luminosities and masses for sources with significant extinction. However, the change in near-infrared reddening is quite small because, as shown by CCM, all extinction laws converge at wavelengths ≳0.9 μm, such that A(λ)/A(I) is similar in all known clusters. Only the optical extinction is greatly modified and we are not concerned with this. Hence we use A(J)=2.783E(J−H), which compares with A(J)=2.364E(J−H) for the R=3.05 interstellar extinction law derived from Whittet (1992) and A(J)=2.543E(J−H) which follows from the oft-quoted λ^−1.8 law for near-infrared interstellar extinction. This convergence of extinction laws in the near-infrared is fortunate, since we doubt whether a common extinction law will apply to all the stars in a given star formation region, especially in Orion where the nebulosity has such complex spatial structure.

3.2 Dereddening procedure

For the IJH detections, we have used the unusual procedure of dereddening to an empirically derived curve in the (I–J) versus (J−H) two-colour diagram, rather than to a theoretical curve in a colour-magnitude diagram. This was for two reasons: first, theoretical models are at an early stage of development for such young substellar sources and appropriate colour predictions have yet to be published at the time of writing; secondly, theoretical colour predictions are subject to the significant uncertainty in the absolute temperature-colour calibration that we referred to in Section 1. The empirical curve (Fig. 3, top panel) was derived from a fit to observations of M and L dwarfs of near-solar metallicity from Leggett et al. (1998) and UKIRT stellar standards of unknown metallicity for (I–J)<1, where there is no known metallicity dependence. A possible flaw in this approach is that young stars and brown dwarfs have lower surface gravities than main-sequence objects of the same effective temperature. However, the colour predictions by Baraffe et al. (1998) for young low-mass stars indicate that much larger changes in log g (>1 dex) have a negligible effect on these colours. The B97 models also indicate that log g at 1 Myr does not approach red giant values for the masses considered here, so the empirical curve is not likely to be far in error. A cubic polynomial was used, with a Gaussian addition to fit the peak at (I–J)≈1.0. The form of the relation is a good match to the plots of Bessell & Brett (1988) and Leggett (1992).

Figure 3.

Top panel: empirical two-colour curve fitted to the plotted observations. Two examples of dereddening tracks are also shown, indicating the possibility of double-valued solutions. Bottom panel: results of dereddening Orion data.

Open in new tabDownload slide

The results for the 272 IJH sources that are not anomalously blue in (I–J) are plotted in Fig. 3 (lower panel). 95 sources have double-valued solutions but only one solution has a plausible colour and flux in nearly every case. The handful of uncertain choices are low-mass stars near the bump in the curve at (I–J)≈1.0, where the two solutions lie close together in (J−H) and derived luminosity but differ substantially in (I–J) and hence derived temperature. These ambiguous sources are listed as such in the data table (available electronically from the authors by FTP). We are confident that the dereddened (J−H) colours are accurate to ±0.1 mag in nearly every case, given the weak temperature dependence of this colour; a standard error of 0.05 mag is estimated, owing to measurement error and uncertainties in the empirical curve and dereddening law. The J-band extinction correction should therefore have typical uncertainties of ±0.14 mag, which is small enough to produce a useful luminosity function. The dereddened (I–J) colours are more sensitive to any errors in the process, such that the standard error is approximately 0.25 mag. However, the very strong temperature dependence of the (I–J) colour means that this leads to only a modest uncertainty in derived values of T_eff (cf. Section 4.3).

The 202 JH-only detections were dereddened to the theoretical track shown in the colour-magnitude diagrams (Fig. 1), which is a simple linear fit to an L–T_eff relation (taking the average of the DM98 and B97 predictions at 1 Myr), and a T_eff-(J−H) relation [taking the average of the Wilking et al (1999) and Baraffe et al. (1998) relations for main-sequence stars, which agree to 0.02 mag]. Only the B97 and Baraffe et al. predictions extend to the faintest magnitudes and lowest temperatures [(I−J)>3.3]. In this region the (J−H) colour changes more rapidly with T_eff and the uncertainties increase.

4 Interpretation

4.1 Luminosity function and initial mass function

The luminosity function (LF) is plotted for all sources detected in the H band with m_H>12.25 in the top panel of Fig. 4. The function declines from a strong peak at m_H⸒.5, which is not well measured here because of saturation. A strong secondary peak at H=16.5 is apparent, which probably has physical significance given that the function is based on more than 500 sources, and any real features are blurred by extinction of typically 1 mag in the H band. The peak exists independently of the magnitude binning and is seen in the right-hand panel of Fig. 1 as a clump of sources with 16<m_H<17. A corresponding feature exists in the J-band LF at J=17.5. The completeness falls gradually, owing to the variable nebular surface brightness, but is estimated as >90 per cent to m_H=18.0, as evidenced by the small peak there. Zinnecker, McCaughrean & Wilking (1993) and Ali & Depoy (1995) observed the equivalent K-band function and found a peak at m_K=11–12, the function declining to fainter magnitudes but flattening off and possibly rising beyond m_K=14. A deeper K-band LF based on the central 5×5 arcmin² region (McCaughrean et al. 1995) shows the second peak at K=15.0 to 15.5, which provided the first indication that a significant population of brown dwarfs might exist in the Trapezium cluster.

Figure 4.

Top panel: observed H-band luminosity function. The equivalent J-band function is overplotted as a dashed line. Both functions are complete to approximately magnitude 18 (dotted line). Bottom panel: dereddened luminosity function, complete between magnitudes 6.2 and 11.75 (dotted lines).

Open in new tabDownload slide

The observed function is converted to the absolute LF shown in the lower panel of Fig. 4, including only the dereddened sources (i.e. excluding H-band-only detections and blue IJH sources). M_bol is determined for each source using J-band magnitudes and bolometric corrections (these being well established at J) and a distance modulus of 8.22. The bias arising from necessarily retaining some faint, anomalously blue sources owing to lack of I-band data is only significant below m_J=18, which is approximately the completeness limit (the nebulosity reduces sensitivity more in the I and J bands than at H). We adopted the following bolometric corrections, derived from a simple fit to the relations between T_eff, (J−H) and BC_J of Wilking et al. (1999) [and Baraffe et al. (1998) for T_eff>3500 K] and using the DM98 1-Myr isochrone to connect T_eff to luminosity:

(3a)

(3b)

where J_dr is the dereddened J magnitude and BC_J refers to the UFTI filter, which has smaller bolometric corrections than the CIT filter.

The lower panel of Fig. 4 shows a primary peak at M_bol=6 and a small secondary peak at M_bol=10.5. Incompleteness is significant for M_bol<6.2, owing to saturation of bright sources, and for M_bol>11.75, which corresponds to J>18. The secondary peak corresponds roughly to the peak at H=16.5 in the top panel of Fig. 4. We convert the LF to the initial mass function (IMF) using the tracks of B97 and DM98. To remove bias owing to non-detection of highly reddened sources we include only sources with (J−H)_obs<1.5. The results in Fig. 5 therefore represent an unbiased sample of the IMF complete to log(M/M_⊙)=−1.5. In a log-log plot, both IMFs show a fairly flat stellar function which has a tendency to fall slightly into the brown dwarf regime. Both IMFs also show some indication of a rise beyond the completeness limit, but deeper observations will be needed to quantify this.

Figure 5.

IMFs for B97 and DM98 tracks, complete to log(M/M_⊙)=−1.5. Error bars are plotted assuming Poisson statistics. The IMF appears to fall slowly into the brown dwarf regime, but rises again below the completeness limit (dotted line).

Open in new tabDownload slide

The discovery of a large population of brown dwarf candidates contrasts with previous surveys (e.g. Hillenbrand & Hartmann 1998) which have concluded that few substellar objects exist. This conclusion appears to have been based on the well-established decline in the LF beyond the principal peak at m_K≈11.5. However, this survey probes deeply enough at infrared wavelengths to detect the secondary peak in the LF.

4.2 A cut-off in the luminosity function?

The absence of faint blue sources (see Section 3) is well established by the inclusion of the H-band upper limits. This may indicate a sharp turn-down in the LF, and perhaps a cut-off, at a level corresponding to about 8 Jupiter masses. However, such sources would be close to the survey sensitivity limit (which we believe to lie just below 5M_Jup), particularly if their intrinsic (J−H) colours are redder than we expect. Moreover, the B97 mass-luminosity relation becomes steeper below about 8M_Jup, so a turn-down in the LF would occur even for a flat IMF. Hence a deeper survey will be needed to confirm the reality of the fall in the IMF. The least massive detection with good photometry is Orion 023-115, which has J=19.38, (J−H)=0.97 and a derived mass of for an age of 1 Myr, using the B97 tracks. The quoted ± uncertainties refer to alternative ages of 2 and 0.3 Myr respectively. None of the faint sources in the left-hand panel of Fig. 1 with highly uncertain J-band fluxes has a lower derived mass. If real, the turn-down might be attributed to a minimum Jeans mass for gravitational cloud core collapse, below which star formation cannot occur. Alternatively, it may be due to the ending of star formation in the cluster before extremely low-mass cloud cores had time to collapse, since this process is believed to take longer in less massive cores. This may be attributable to dispersal of dense molecular gas by the photoionizing O-type stars at the centre of the cluster. The IAC group (Bejar et al., in preparation) apparently find no sign of the turn-down in the LF at planetary masses in the neighbouring σ Orionis cluster, so we favour the second explanation. If correct, this explanation may still be significant with regard to any Galactic population of free-floating planets: most star formation is believed to occur in high-mass star formation regions like the Trapezium or M16, with the consequence that free-floating planets may be relatively rare. However, the effectiveness of this mechanism for reducing planet formation efficiency will vary depending on local conditions.

4.3 Effective temperatures

The dereddened (I-J) colours are converted to the effective temperatures listed in the data table (available by FTP from the authors) using the models of Baraffe et al. (1998). The derived values of T_eff are in reasonable agreement with the predictions of B97 and DM98, in which 2600<T_eff<2900 K for objects of brown dwarf mass at 1 Myr. However a few sources in this mass range are detected with dereddened (I–J) ⩾ 3, which implies T_eff ⩽ 2500 K, at least for main-sequence objects. We estimate that our derived temperatures are accurate to ±200 K, leaving aside the uncertainty in the absolute T_eff versus (I–J) relation and assuming that this is not altered significantly by the youth of the sources. In any case, the (I–J) colours should provide a useful guide to relative temperatures.

4.4 Potential problems

We have carefully avoided several potential problems in studies of young clusters, such as cluster membership, infrared excess and line emission, but some serious issues remain. We consider each in turn.

• (i)

  Foreground contamination is minimal in the tiny area surveyed, introducing perhaps five red dwarfs into the sample over a range of about 3 mag, using the stellar space densities of Tinney (1993).

• (ii)

  Background contamination is removed by the dark backdrop of OMC-1 (see Hillenbrand & Hartmann 1998). Our avoidance of the K-band filter removes any chance of seeing through OMC-1 at the faint limits.

• (iii)

  Infrared excess arising from hot dust is not believed to be significant in the H band, since the spectra of T Tauri stars are well fitted by blackbodies at this wavelength.

• (iv)

  Line emission was minimized by the choice of filters.

• (v)

  Variation in the infrared extinction law will probably be small (see Section 3.1) and the effect on the derived IMFs is minimized by excluding highly reddened sources.

• (vi)

  Scattering is a potentially serious problem. Even sources without anomalous colours in the I versus (I–J) diagram may have distorted (J−H) colours and fluxes. As noted in Section 3, however, photospheric flux dominates the spectral energy distributions of most T Tauri stars for 0.6<λ<2 μm, so significant distortion of broad-band colours and fluxes is unlikely to be common. This should be investigated in future by searching for the polarization signature of scattered light.

• (vii)

  The evolutionary tracks for substellar objects are in an early stage of development, which leads to a significant uncertainty in derived masses. However, the fairly close similarity of the B97 and DM 98 tracks, for luminosity and T_eff, is encouraging.

5 Conclusions

A large population of brown dwarf candidates is detected in the Trapezium cluster, along with a small population of objects with planetary masses. We have confidence that these are true cluster members and, although many uncertainties exist in deriving the masses, they are not likely to be large enough to cause misclassification of low-mass stars as low-mass brown dwarfs or free-floating planets. The derived IMF is fairly flat on a log-log plot at low stellar masses, but declines slightly at brown dwarf masses, indicating that brown dwarfs are a little less numerous than stars. There is a possible small peak near ∼0.02 M_⊙, which is below the completeness limit. Approximately 13 planetary-mass objects are detected, but none with M<8×10⁻³ M_⊙. We suggest that this is due to dispersal of the star-forming cloud by the photoionizing O stars before such objects had time to form.

Approximately 13 per cent of sources have anomalously blue (I–J) colours, which we attribute to scattering from circumstellar material. These blue excesses are strongly correlated with detection as ‘proplyds’ by HST, so this colour selection may prove to be a powerful new tool for detecting sources with circumstellar envelopes. The effect of scattering on the colours of the general population should be investigated via polarimetry and spectroscopy.

Acknowledgments

We thank the staff of UKIRT, which is operated by the Joint Astronomy Centre on behalf of the UK Particle Physics and Astronomy Research Council (PPARC). Particular thanks are due to Sandy Leggett and Andy Adamson for providing information on the colours of cool stars and for carrying out reactive observations for us. We also thank the Panel for the Allocation of Telescope Time for providing us with reactively rescheduled observing time. We thank Juliette White for helping us with the dereddening procedure, and we are grateful to the referee, Richard Jameson, for useful and timely comments. PWL is grateful for support by PPARC via a Post Doctoral Fellowship at the University of Hertfordshire.

References