A summary of the best fitting parameters for each source is given in Table 5. The quality of the fit was estimated from the average of the quadratic sum of the residuals, most of them weighted with the inverse of the estimated error in magnitude. We gave only half of this weight to the residuals in R and I, because the combination of the cool temperatures of our objects and the strong extinction cause large flux gradients in this region, and the effective extinction over the filter width thus becomes a complicated function of the spectral energy distribution of the source and the transmission curve of the filter. A smaller weight was also given to the measurements longwards from SW1, due to the fact that our modeling of the intrinsic infrared excess probably gives only a rough approximation to the true spectral energy distribution in the disk emission-dominated region
Table 5. Best fitting parameters to the available photometry
In Table 5, we give for each object the visual extinction (assuming ; Rieke & Lebofsky 1985), spectral index n, temperature T, and mass M. Also indicated is whether an "old" or a "young" isochrone provide the best fit, where the separation between these cases is set at years. This age is about half of the estimated age of the Ophiuchi complex given by Wilking et al. 1989, which is roughly consistent with that estimated by Comerón et al. 1996 in their reanalysis of the CRBR data. Although the isochrones considered were spaced by intervals of years, the quality of the fits was insensitive to changes between consecutive isochrones; therefore, we feel that it is not meaningful to give more precise values of the best fitting age. The mass given by the fit, on the contrary, does have some sensitivity to the chosen isochrone, especially at the earliest evolutionary stages, and choice of one or another particular isochrone within the same age group may occasionally move a star from the brown dwarf to the stellar domain or vice versa. The masses listed in Table 5 correspond to ages of years for the young group, and years in the old group. In some cases for which the quality of the fit is similar for ages in the "young" or the "old" range, it has been possible to make a choice based on spectra obtained by Williams et al. 1995 and, more recently, by Meyer et al. 1998. Such cases are noted in the discussion of individual objects below.
The best fitting spectral energy distributions given by the models for the objects listed in Table 5 are illustrated in Fig. 1. Also plotted are the contributions of the stellar photosphere without circumstellar excess in the cases when such excess is required to obtain a good fit.
The required circumstellar excess is fairly insensitive to the chosen isochrone over the whole range of ages considered here. However, even moderate deviations (by a few tenths) in the slope of the excess, n, from the values quoted in Table 5 produce clear departures from the observed fluxes at the longest wavelengths. These deviations cannot be removed by varying the level of foreground extinction or the temperature of the central object. The constraints on n are particularly strong for the objects detected at LW1 or LW4, which stresses the importance of the good spectral coverage provided by ISO in improving the object/excess/extinction decomposition. To illustrate this point, Fig. 2 shows the best fits to 2320.8-1721 at wavelengths longer than 1.25 µm obtained by imposing three different values of the spectral index: (the best fitting one; solid line), (dotted line), and (dashed line). In each case the extinction and photospheric temperature have been optimized. It is clear that only produces an acceptable fit to the shortest and the longest wavelengths simultaneously, even if the fluxes longward from LW1 suggest that more excess may be required to fit this region. However, a better model fit at long wavelengths would produce discrepances exceeding a factor of 2 in the photosphere-dominated near infrared, where the JHK fluxes are generally measured with a greater accuracy and the spectral energy distribution modeling is much more reliable. Our fits give more weight to the short wavelength measurements, as explained above; departures at the longer wavelengths are most probably due to our simplified way of modeling the excess of circumstellar origin, coupled with the smaller photometric accuracy.
4.1. Discussion of individual objects
The ISOCAM measurements are poorly fitted 1, but the detection at LW1 and LW4 suggest a cold infrared excess. In fact, the large LW4 flux suggests a spectral index shallower (i.e., with a smaller absolute value) than -2.2 which, if extrapolated to shorter wavelengths, would imply a smaller photospheric contribution at K and an accordingly lower mass and luminosity.
2320.0-1915 = GY5
2320.8-1708 = GY10
2320.8-1721 = GY11
Our best fit is obtained with a moderate age ( years) and a spectral index , yielding . The slow early evolution of objects with such a mass makes this fit nearly mass-independent within the range of ages considered here. A fit implying a mass greater than 0.08 would be possible only by assuming an age greater than years, which seems to be ruled out by the estimated age of the Ophiuchi complex (de Geus 1992). For this source, such an age is particularly unlikely because of the well-determined infrared excess (see Fig. 1), which would be expected to be uncommon in objects older than years. Moreover, a fit with an age in excess of years would require a temperature above 3000 K, in disagreement with the feature strengths in the 2 µm spectrum.
Despite its youth and the evidence for circumstellar matter around it, the possibility of further accretion raising the mass of this object much above its present value seems very unlikely in view of its non-detection at 1.3 mm by André & Montmerle 1994.
An intriguing aspect of this object is the large discrepancy in the measured values of H between CRBR and Strom et al. 1995, amounting to 1.15 magnitudes; however, the difference in K is only 0.27 magnitudes, in agreement within the uncertainties. Follow-up observations should be carried out to confirm such apparent extreme variations in color.
2331.1-1952 = GY64
This object is promising for further spectroscopic followup. It is only moderately obscured, like 2320.8-1721, making it a promising target for follow-up spectroscopic observations in the visible. At longer wavelengths, on the other hand, the small intrinsic infrared excess should decrease the importance of veiling which otherwise complicates the interpretation of spectra in the 2 µm region (Luhman & Rieke 1998). Its spectrum in this region has been recently obtained by Meyer et al. 1998, and the spectral type M8.5 derived by these authors strongly support the brown dwarf character.
2349.8-2601 = GY141
A considerable infrared excess, , is required to produce an overall fit to the available photometry, although with rather large residuals over the JHK bands. The non-detection at LW4 is still consistent with such an excess. The temperature derived from the fit, 2450K, compares well with that measured spectroscopically by Luhman et al. (1997), 2700 150K. The fitted mass is for an age below years, again in reasonably good agreement with the estimate of 0.045 0.015 derived by Luhman et al. (1997) by placing the object on the HR diagram and using the Burrows tracks as we have.
2351.8-2553 = GY146
2404.5-2152 = GY202
2408.6-2229 = GY218
4.2. Comparison with other work
The well-constrained fits made possible by combining groundbased and ISOCAM photometry make it of interest to compare the temperatures derived by isochrone fitting with spectral determinations. Details are given in the discussion of individual sources. Of seven objects measured with both techniques, extending from 0.2 down to 0.03 , excellent agreement is achieved in all cases. In addition, although all differences are within the expected errors, there is also no discernible trend - that is, the residuals show no bias toward high or low temperatures from the fitting technique compared with spectroscopy. A noteworthy comparison is 2349.8-2601, whose optical spectrum is analyzed in detail by Luhman et al. (1997). Both the temperature and the substellar mass they derive agree with our values from isochrone fitting.
Given the good agreement between spectroscopy and the broadband fits, we can test the quality of the fits based on groundbased infrared photometry alone. Fig. 3 compares the masses derived for the objects studied here with those of CRBR. A good overall agreement exists, despite the more limited wavelength coverage of CRBR and their use of older models of stellar interiors and atmospheres. The only object for which the mass estimate is significantly changed is 2317.5-1729, for which CRBR assumed a strong circumstellar excess unconfirmed by the ISOCAM observations in LW1 and LW4 (see discussion above). The present results thus support the usefulness of fits to ground-based photometry, or even JHK only, to study the mass function of embedded populations. At the same time, confirmation of the substellar nature of objects with masses suspected to be near the brown dwarf limit but with strong excesses appears to require observations over a wide range of wavelengths (or spectroscopy).
In the alternate method of Strom et al. (1995), the JHK color-color diagram is used to deredden the sources and estimate their J luminosities, which are finally compared with the theoretical isochrones to obtain masses. The results have been compared with those of CRBR by Williams et al. (1995). The agreement between the two approaches is very good. Thus, the validity of their method is also supported by the ISOCAM observations and analysis.
© European Southern Observatory (ESO) 1998
Online publication: June 18, 1998