2 Observations and data reduction

All of our CCD survey was carried out during 1996 in two campaigns which took place on February 9-12 and on September 19-20. We used the TEK (1024$\times$1024 pixel²) detector mounted on the prime focus of the telescope, with a field of view of 10$\times$10arcmin². A total of 40 fields, with center coordinates listed in Table 1, were observed with the Harris (R)I and RGO Z broad-band filters providing a total survey area of 1.05deg² ($\sim$6.5% of the whole cluster). Most of these fields were selected to avoid very bright stars and were located within 1deg of the innermost region of the Pleiades (see Table 1), where the population of M dwarf proper motion members is much larger than in outer areas (Hambly et al. 1993; Bouvier et al. 1998). If as expected less massive Pleiades members show a similar spatial distribution within the cluster our survey should be able to detect a large number of BD candidates. In Fig. 1 the location of all the frames obtained during the two observing runs is presented to scale. None of them (except for five fields indicated in the figure) falls within the small region southwest of the cluster center which is well known to suffer from high absorption (van Leeuwen 1983; Breger 1987; Stauffer & Hartmann 1987). A small fraction (10.5%) of our covered area was imaged in the three bands while a similar area was observed only in RI filters. Weather conditions were always photometric, and the seeing oscillated between 1$^{\prime\prime}$ and 1$.\!\!^{\prime\prime}$5. Typical exposure times ranged from 10min for the R filter to 5min for the I and Z-bands.

 

Table 1:   Field (100arcmin²) center coordinates

$^{\rm{a}}$

Separation from the cluster center (3$^{\rm{h}}$ 47$^{\rm{m}}$, +24$^\circ$7$^\prime$).

 

Figure 1:   Location of our fields (squares) within $3\hbox{$.\!\!^\circ$}5 \times 3\hbox{$.\!\!^\circ$}5$ of the Pleiades area. Open squares stand for those fields observed with IZ filters while the shaded squares depict the four fields observed with RI filters. Central coordinates (indicated with a cross) are 3$^{\rm{h}}$ 47$^{\rm{m}}$, +24$^\circ$ 7' (Eq. 2000). The five fields which may have some amount of extra reddening according to the CO contours shown by Breger (1987) are indicated with an arrow. Filled star symbols stand for stars brighter than 6$^{\rm th}$magnitude, and filled circles for proper motion M members (Hambly et al. 1993) with I magnitudes in the range 13-18. The vertical gap in the M star distribution around 3$^{\rm{h}}$ 51$^{\rm{m}}$ is due to the fact that there were no overlaps between the first and second epoch plates used by the authors, causing the lack of proper motion measurements for stars in that strip. The relative brightness is represented by symbol diameters. North is up and East is left

We adopted the IZ broad-band filters for several reasons. One of our goals was to detect objects fainter and less massive than the two cluster BDs Teide1 and Calar3 (M8, $I\,\sim\,19$, $R-I\,\sim\,2.6$, $\sim\,0.055\,M_{\odot}$, Rebolo et al. 1995; Martín et al. 1996). Theoretical evolutionary models (which do not include grain formation in very cool atmospheres) predict that these objects become much redder with colours (R-I) $\ge$ 3 (Chabrier et al. 1996). Thus, they might be extremely faint in R wavelengths, greatly hindering their detection. On the other hand, field stars do exhibit a turn-off in (R-I) at around M7 spectral type, with stars of later types having bluer colours (Bessell 1991). The fluxes and colours of the Pleiades BDs fainter than Teide1 and Calar3 are unknown, but we expect them to have spectral energy distributions which resemble those of the coolest objects in the field. It could turn out that the (R-I) colour is no longer useful to discriminate low luminosity cluster members from field objects. The (I-J) colour, however, gets monotonically redder for lower temperatures (both for observed and theoretical predictions), implying that the slope of the spectral pseudocontinuum between I and J wavelengths clearly increases. As the Z filter is centered at 920nm, we expect a similar behaviour with I and Z. Although the efficiency of the CCD drops considerably in the Z-band, this effect is compensated by the increased brightness of BDs at these near-IR wavelengths. The (I-Z) colour has been shown to be a useful discriminant for Pleiades BDs by Cossburn et al. (1997).

Other photometric searches for substellar objects in the Pleiades carried out with R and I (Jameson & Skillen 1989; Zapatero Osorio et al. 1997b, PaperI) provide a high number of mid- and late-M stars that do not belong to the cluster and are contaminating the surveys. It is desiderable to find a strategy which avoids these field contaminants and facilitates a more efficient tool for detecting true members. In PaperI the success rate was only 25%: two out of the eight proposed cool candidates have been confirmed as genuine Pleiades BDs (Rebolo et al. 1996). The authors argue that this was due to the detection of reddened late-M dwarfs (Zapatero Osorio et al. 1997c, PaperII). The use of longer wavelength filters would help to jump over this obstacle.

 

Figure 2:   Photometric errors as a function of observed I (full line) and Z (dots) magnitudes. The observed dispersion in the errors for the I-band is similar to that of the Z-band

Raw frames were processed using standard techniques within the IRAF (Image Reduction and Analysis Facility) environment, which included bias subtraction, flat-fielding and correction for bad pixels by interpolation with values from the nearest-neighbour pixels. The photometric PSF fitting analysis was carried out using routines within DAOPHOT, which provides image profile information needed to discriminate between stars and galaxies. Instrumental RI magnitudes were corrected for atmospheric extinction and transformed into the RI Cousins system using observations of standard stars from Landolt's (1992) list. Special care was taken in including red standard stars in order to ensure a reliable transformation for the reddest candidates: the field SA98 contains many photometric standards covering colours from A0 to M7 spectral type. The calibration of Z magnitudes required more observational effort as there are no real data for standards available in the literature. We have not performed an absolute flux calibration for this filter, but obtained (I-Z) colours with respect to a given spectral type. Using the same Landolt fields as observed through the other two filters at culmination (airmass = 1.1), we set Z = I for those standard stars with (R-I) $\sim$ 0 (A0-type). The adopted (I-Z) colours are shown in Table 2. Observations of these fields at different elevations allowed us to correct Z instrumental magnitudes for atmospheric extinction. Errors for Z instrumental magnitudes as provided by IRAF routines are plotted in Fig. 2. The best power law fit to the errors in I for the bulk of data is superimposed in the figure for comparison. Summarizing, uncertainties in the INT photometry range from $\le$0.05mag at I, Z $\sim\,20.5$, 19.7 to about 0.15mag at 22, 21mag, respectively.

 

Table 2:   Adopted (I-Z) colours for the photometric standard stars used in the calibrations

^a

RI magnitudes and their errors taken from Landolt (1992).

^b

1-$\sigma$ errors come from uncertainties in I-band and the dispersion in the calibration.

We present in Fig. 3 the resulting I vs. (I-Z) diagram where data for the Pleiads HHJ3, PPl15 and Teide1 (which are present in three of our fields) are combined with the new observations. We remark that Z magnitudes are not on a standard system. Completeness and limiting magnitudes of our survey were derived following the same procedure described in Stauffer et al. (1994) and PaperI. We estimate them to be I, $Z \sim\,21$, 20.5 for completeness and 22.2, 21.5 for the limit. These values are indicated in the figure. Because there are almost no measurements in the Z-band of other cluster members, it is rather difficult to establish the separation between Pleiads and field objects in our diagram. However, we have made an attempt to separate these two kinds of objects by plotting a straight line in Fig. 3 which is parallel to the photometric sequence defined by HHJ3, PPl15 and Teide1 and shifted 0.3mag towards the blue. For bluer colours than those indicated by the line, the number of detections increases very drastically, probably indicating that these are field objects. On the other hand, the photometric dispersion observed in other optical and infrared colours among low-mass proper motion members is about 0.6 - 0.7mag (Steele & Jameson 1995; Martín et al. 1996). Given the proximity in wavelength of the I and Z filters it is expected that this dispersion becomes smaller and therefore, the adopted shift takes into account possible binarity effects. For example, PPl15 was first claimed to be a photometric binary in PaperII and actually it has been confirmed as a double-lined spectroscopic binary with nearly identical components (Basri & Martín 1998). Those objects fainter than HHJ3 and PPl15 and located on the right side of the straight line are considered our best BD candidates. There are 43 BD candidate members of the Pleiades in total, plus one (slightly brighter) very low-mass candidate stellar member of the cluster. A better definition of the true location of this line should be derived after IR photometry and spectroscopy are obtained for the candidates (Zapatero Osorio et al. 1998b).

 

Figure 3:   IZ colour-magnitude diagram for our 1.05deg2 survey in the Pleiades. Z magnitudes are not on a standard system (see text for details). Previous known members are labelled along with the completeness (dashed line) and limiting (full horizontal line) magnitudes. Suspected extended objects are shown with asterisks, and the seven candidates previously studied in Zapatero Osorio et al. (1997a) are indicated with filled triangles. Masses according to the NG Chabrier et al.'s (1996) model for solar metallicity and 120Myr are labelled on the right side

 

Table 3:   Coordinates and photometry for the candidates

$^{\rm{a}}$

"RPL'' stands for Roque Pleiades, and "TPL'' for Teide Pleiades.

$^{\rm{b}}$

References: PPl objects from Stauffer et al. (1989); JS objects from Jameson & Skillen (1989); NPL objects from Festin (1998); CFHT-PL objects from Bouvier et al. (1998).

$^{\rm{c}}$

Coordinates for HHJ3 taken from Hambly et al. (1993); for PPl15 from Stauffer et al. (1994); coordinates and RI photometry for Teide1 taken from PaperI.

$^{\rm{d}}$

These objects appear slightly extended in the IZ images.

:

For error bars in the photometry see text. Those measurements labelled with a ":'' have rather large uncertainties.

In Table 3 we list the names, magnitudes, colours and positions for the proposed Pleiades BD candidates. They are named after the Roque Observatory followed by the word Pleiades and numbered according to their decreasing I-band apparent magnitude (second column of Table 3). Hereafter, we will use an abridged version of the names which omits the term "Pleiades''. The names of the candidates adopting the IAU rules are also provided (first column), where the acronym "RPL'' stands for Roque Pleiades. Three of the four faintest candidates have slightly larger fwhm than the average value for our frames. Presumably this is an indication that they are not a point source. It is expected that distant galaxies fainter than I = 21 will begin to contaminate the number counts of objects. Those candidates labelled as extended are shown with a different symbol in Fig. 3. By reference to the reddening map provided in Breger (1987), Roque3, 5, 15, 18 and 32 could suffer from a somewhat enhanced extinction as they lay within or very near to the CO contours given by the author.

In addition to the INT data, we have obtained R-band photometry for five of the candidates at the 2.5m Nordic Optical Telescope (BroCam1, NOT) on 1996 October 10-11 (Roque17, 11 and 4), and at the 1m Jacobus Kaptein Telescope (JKT) on 1996 September 12-13 (Roque16 and 13), both telescopes at the ORM. The CCDs used were a Tektronix 1024$\times$1024 providing fields of view of 3.0 and 5.5arcmin², respectively. Exposure times were typically 15min at NOT and 30min at JKT. Landolt's (1992) standard stars were observed just before and after the targets. Reduction of the raw frames and photometry of the candidates has been performed as described above. Uncertainties in R magnitudes range from 0.07mag for the brightest objects to 0.15mag for the faintest ones. Considering the R-I photometry from Table 3 and from other deep surveys (PaperI; Bouvier et al. 1998) Roque44 and Roque26 are not likely to be Pleiades members as they seem to deviate towards bluer colours from the sequence defined by other candidates.

Astrometry for all Roque objects has been performed by the triangles fitting method using the APM Sky Catalogue. Several stars close to every candidate were identified and they served as a reference for the astrometric calibration. Coordinates are accurate to approximately $\pm$2$^{\prime\prime}$. The location of our candidates in the surveyed area is depicted in Fig. 4. Their distribution around the cluster center appears quite homogeneous. However, we note that the number of fields (9) with $\geq$3 BD candidates is surprisingly large compared to the expectations from a random distribution. The study of possible spatial inhomogeneities within the cluster still awaits membership confirmation. Seven of the Roque BD candidates have also been identified in other surveys. The last column of Table 3 gives cross-identifications. Our I magnitudes seem to be on average 0.25mag brighter than those available in the most recent literature. This is likely due to an effect of the colour-dependence of the Harris filters we used in our observations; although a red standard star was considered, it is poorly calibrated and consequently does not provide an accurate determination of the colour-term in the photometric calibration. Cossburn et al. (1998) have found that the colour-term for the transformation from Harris I to Cousins is indeed rather significant for very red objects. In the case of Roque33 (NPL40) the difference found is -0.58mag which might be due to contamination from a nearby very bright star (and saturated in our frames). Figure 5 provides the I-band finder charts (2$^\prime$$\times$ 2$^\prime$ in extent) for all Roque objects ordered as listed in Table 3.

 

Figure 4:   Location of our candidates within the fields observed in our survey covering 1.05deg2. As in Fig. 1 central coordinates (3$^{\rm{h}}$ 47$^{\rm{m}}$, +24$^\circ$ 7', Eq. 2000) are indicated with a cross. Filled star symbols outline stars brighter than 6$^{\rm th}$magnitude. The relative brightness is represented by symbol diameters. North is up and East is left

According to the "NextGen'' (NG) theoretical evolutionary models of Chabrier et al. (1996), and adopting solar metallicity, an age of 120Myr (Basri et al. 1996; Martín et al. 1998; Stauffer et al. 1998) and a distance of 127pc for the Pleiades cluster, our survey has detected objects in the mass interval from roughly 0.08$M_{\odot}$ down to 0.03$M_{\odot}$. The completeness magnitudes correspond to 0.035$M_{\odot}$ as indicated in Fig. 3. Chabrier et al.'s models provide absolute magnitudes as a function of mass, metallicity and age obtained by direct integration of theoretical atmospheres which do not incorporate grain formation and dust absorption (Allard et al. 1997). However, the effects of condensation become important for temperatures cooler than about 2500K (Tsuji et al. 1996; Jones & Tsuji 1997), a temperature range partially covered by our survey. Preliminary computations by Baraffe (private communication) show that models considering dust formation and opacities predict brighter I magnitudes and subsequently slightly lowers the mass determination by $\sim8$%.

Membership and therefore the real nature of our candidates on the basis of JHK photometry and spectroscopy and the Pleiades mass function will be addressed in a forthcoming paper (Zapatero Osorio et al. 1998b). Seven of them (Roque17, 16, 15, 14, 13, 11 and 4) with I magnitudes in the range 17.8 - 19.5 (masses in the interval $0.08-0.045\,M_{\odot}$) have already been studied to some extent by Zapatero Osorio et al. (1997a). They are shown in Fig. 3 with a different symbol. The authors conclude that given their K magnitudes, radial velocities, spectral types and weakness of some atomic features these candidates should be considered as Pleiades members. The number of remaining candidates in our IZ survey deserve further investigation as there are large enough to ensure that follow-up observations will confirm more Pleiades substellar objects. Among the faintest ones, there could be BDs with masses as low as 0.03$M_{\odot}$. These studies will make it possible to derive the cluster mass function well into the substellar regime.