1 Introduction

The DEep Near-Infrared Survey (DENIS) will be a complete near infrared survey of the southern sky (Epchtein et al. 1994; Epchtein 1997). It aims to provide full coverage in two near-infrared bands (J and $K_{\rm s}$) and one optical band (I), using a ground-based telescope and digital array detectors. The approximate $3-\sigma$ limits of the survey are I=18, J=16, K=13.5. The products of this survey will be calibrated images and catalogs of point sources and extended sources. The survey started in January 1996 and is expected to be completed within five years.

Two dedicated data analysis centres (DACs) have been created to process this data. The Paris DAC processes the raw data from the telescope into flattened and cleaned images (Borsenberger 1997). The Leiden DAC will then process these images to extract point source information (Deul et al. 1995), which will be assembled into a database at the Paris DAC.

One area of astronomical research in which such a database will have a particularly profound impact is the study of very low-mass stars (VLMs) and brown dwarfs (BDs). Prior to the current generation of all-sky infrared surveys, the two least luminous star/brown dwarf objects known - Gl229B (Nakajima et al. 1995) and GD165B (Becklin & Zuckerman 1988) - were both detected as companions to brighter nearby stars. While such "looking for things around other things'' techniques improve the chances of finding rare objects, they do not allow us to determine their space density. Younger and warmer free floating brown dwarfs have also been identified in the Pleiades cluster (Rebolo et al. 1995; Basri et al. 1996). The most recent Pleiades work (Zapatero Osorio et al. 1997; 1998) suggests a mass function which rises into the brown dwarfs regime as ${\rm d}N/{\rm d}M{\sim}M^{-1}$ (Martín et al. 1998). VLMs and BDs may be thus a numerous, if not dynamically significant, galactic disk population. The space density of the VLM/BDs is thus of fundamental importance to star formation theory. However, because the initial mass function in a particular young cluster may not be representative of the disk altogether (since segregation and cluster evaporation are mass sensitive processes), it is important to determine mass functions for the overall field population.

To date, studies of these populations in the field have been severely hampered by the difficulty of finding such objects. The most sensitive (i.e. largest volume) surveys to date have been carried out using R and I band photographic plates from the Palomar and United Kingdom 48$^{\prime\prime}$-class Schmidt telescopes (e.g. Tinney 1992; Hawkins & Bessell 1988; Gilmore et al. 1985; Reid & Gilmore 1984). However, for the latest-type objects, such searches are limited by the R plate sensitivity and are only complete to $I\approx 16.5-17$. Typically less than one M9-10 dwarf is found in each $6\hbox{$^\circ$}\times 6\hbox{$^\circ$}$ Schmidt field. With a completeness limit of $I\approx 18$, and coverage of the whole southern sky, DENIS therefore offers the chance to detect many more examples of this poorly understood class of object. Moreover, the optical-infrared I-J and I-K colours provided by DENIS are well suited to selecting very cool stars and brown dwarfs (Allard 1998; Burrows et al. 1997).

DENIS will sample very large volumes. The maximum distance for the detection of GD165B-like objects is 20pc, meaning the space density for such objects can be measured to $3-\sigma$ upper limits of $1.8\ 10^{-4}\,{\rm pc}^{-3}$. DENIS is less sensitive to Gl229B-like objects - the limiting distance set by the I-band is 2.3pc. However because the two infrared bands are observed simultaneously with the optical, upper limits are significant, increasing the distance limit to 6.3pc, and the $3-\sigma$ upper limit to $5.7\ 10^{-3}\,{\rm pc}^{-3}$. A constant extrapolation of the observed low-mass stellar luminosity function would predict that DENIS will detect several hundred GD165B-like objects, a few Gl229B-like objects, and many objects (and much needed new data) on objects with intermediate luminosities (Kirkpatrick et al. 1997a). In particular, observation of these new brown dwarfs will address issues fundamental to our understanding of very low temperature atmospheres, such as the importance of dust formation (Allard 1998; Jones & Tsuji 1997; Tsuji et al. 1996a,b), and the importance of previously unrecognized opacity sources like CH₄ (Tsuji et al. 1995) and CrH (Kirkpatrick et al. 1998).

However, to address these issues, significant follow-up programs will be required. With $\pm\,0.3$ magnitude photometry at the survey limits, the DENIS catalogues by themselves cannot be used to construct luminosity functions: first, samples of nearby late-type objects will to some extent be contaminated by distant late-type giants; second, as the number of stars in the catalogs is a steep function of colour (i.e. there are many fewer red stars than blue ones), the number of stars in any one colour (or equivalently luminosity) bin, will be contaminated by stars scattered in from bluer bins; and lastly, cosmic rays and intermittent bad pixels can create spuriously red objects if they happen to overlap an otherwise normal star in one of the infrared pass-bands. In other words, the raw catalog of VLM stars and brown dwarfs produced by DENIS will be a "dirty'' sample. What is this contamination level, and what type of observations will most efficiently "clean-up'' the DENIS catalogue? This is the major issue we seek to address in this paper.