In the past many different criteria have been used to classify a galaxy as a BCD or not. These criteria were commonly based on the galaxy's luminosity and its morphological properties (Zwicky & Zwicky 1971; Thuan & Martin 1981) although definitions based on their spectroscopic properties are also found in the literature (Gallego et al. 1996). Moreover, galaxies morphologically classified as BCDs are sometimes confused with spectroscopically-classified objects, such as "isolated extragalactic HII regions" (Sargent & Searle 1970), "HII galaxies" (Terlevich et al. 1991) or "Sargent-Searle objects" (SS; Salzer, MacAlpine, & Boroson 1989). For instance, although the "isolated extragalactic HII regions" of Sargent & Searle (1970) and SS objects of Salzer et al. (1989) can be undoubtedly classified as BCDs, many HII galaxies in Terlevich et al. (1991) are significantly brighter than a BCD.
The original definition of "Compact Galaxy" comes from Zwicky (1970) where he defined as "compact" any galaxy (or any part of a galaxy) whose surface brightness is brighter than 20 mag/arcsec2 in any chosen wavelength range. The term "blue", as used by Zwicky, refers to those galaxies satisfying the previous condition on both blue and red plates (Zwicky & Zwicky 1971). Later on, Thuan & Martin (1981) introduced the term "Blue Compact Dwarf" referring to those galaxies having absolute blue magnitudes fainter than MB = -18.15 mag (H0 = 70 km s-1 Mpc-1), diameters less than 1 kpc, and strong emission-lines superposed on a blue continuum. More recently, Gallego et al. (1996) spectroscopically classified as BCDs those galaxies showing intense, high-excitation emission lines and low H luminosity (LH < 1041 ergs-1; for H0 = 70 km s-1 Mpc-1). Finally, some variations on these definitions can also be found in the literature for selecting samples of BCD galaxies (Doublier et al. 1997; Kong & Cheng 2002).
In this paper we attempt to unify the concept of BCD by putting forward a new set of quantitative classification criteria. Using these criteria we are able to better include within the BCD class galaxies sharing common physical properties and evolutionary status and segregate them from other types of objects like dwarf irregulars (dIrr) and dwarf ellipticals (dE) or more massive star-forming galaxies. These criteria are sufficiently inclusive so as to also recover most of the galaxies traditionally classified as BCD.
Blue. Probably the most ill-defined property of the "Blue Compact Dwarf galaxies" is the color. Although the presence of a blue continuum in the optical spectra was already required by Thuan & Martin (1981) in their definition of BCDs, it was only a qualitative criterion. The observational criterion traditionally used has been the color of the highest surface brightness component since it was generally the only component detected in surveys using photographic plates. Thus, in order to establish a more quantitative criteria in this same sense we have determined the peak surface brightness (PSB) and the color at this peak surface-brightness from the surface-brightness profiles (Paper II) of the galaxies in our sample, both corrected for Galactic extinction. In order to reduce the effects of different seeing between the images we have averaged the color within the inner 3 arcsec of the profile; in those galaxies with optical diameter 20 arcsec we averaged the inner 1 arcsec.
Figure 1a shows the distribution in color at the PSB measured from the galaxies' surface-brightness. We have also plotted the colors at the PSB obtained for a sample of dIrr (Parodi, Barazza, & Binggeli 2002) and dE (Jerjen, Binggeli, & Freeman 2000) with published surface-brightness and (B - R) color profiles. This figure shows a clear segregation between the color at the PSB of galaxies previous classified as BCDs and that of dIrr and dE, although an overlap is present. We have decided to impose a limit of (B - R) color at the peak of µB, peak - µR, peak 1. Using only this criterion many dE and most of the dIrr galaxies would be classified as BCDs. However, the combination with other criteria will improve the situation significantly (see below). Even with this wide limit some galaxies traditionally classified as BCD by other authors are apparently quite red. Particularly noticeable are the cases of UCM 0049-0045, UCM 1446+2312, and VCC 0001, that show peak color redder than µB, peak - µR, peak = 1.2 and should not be classified as BCDs.
Figure 1. a) Frequency histogram of the (B - R) color at the peak of the surface-brightness profile for our sample of BCD galaxies and a sample of dIrr (Parodi et al. 2002) and dE (Jerjen et al. 2000) galaxies. An average galactic-extinction correction of AB = 0.1mag has been applied to the reference samples data. b) The same for the peak surface brightness. The values obtained for a sample of dIrr galaxies from Patterson & Thuan (1996) are also shown. c) (B - K)-(B - R) color-color diagram for the 21 galaxies in our Atlas with 2MASS K-band magnitudes available. The predictions of evolutionary synthesis models are shown (see text for details). The thick, solid-line represents the best fit to the data. d) Frequency histogram of derived absolute K-band magnitudes obtained applying the relation between the (B - K) and (B - R) colors shown in Panel c.
Compact. With regard their compactness, Thuan & Martin (1981) set a upper limit to the optical diameter of BCDs of 1 kpc. However, observations carried out with CCDs during the 90's have shown the presence of a very extended (up to a few kpc) low-surface-brightness component in many objects that were previously classified as BCDs. Therefore, the term "compact" in BCDs should be related more with the size of the high-surface-brigthness component than with the total optical size. In other words, the compactness criterion should be more a surface brightness limit, like that used in Zwicky (1970), than a real physical-size limit.
In Figure 1b we show the distribution of peak surface brightness (PSB) measured from the galaxies B-band surface brightness profiles (to be presented in Paper II). This figure shows the PSB of BCDs being significantly brighter than that of dIrr (Parodi et al. 2002; Patterson & Thuan 1996) and dE (Jerjen et al. 2000). A fairly good separation between the different types is achieved imposing a limit of µB, peak < 22 mag/arcsec2 to the PSB of BCDs. A total of 9 galaxies in our sample show PSB fainter than this value, so they should not be classified as BCDs. However, an individualized analysis of these objects show that 6 of them are 20 arcsec in diameter that makes the atmospheric seeing to significantly dim their PSB. Five of these objects (HS 0822+3542, UM 382, UM 417, SBS 0940+544C, NGC 4861) also show cometary morphology. Since the PSB have been determined from the azimuthally-averaged surface-brightness profiles, the PSB in these cases is measuring surface brightness at the peak of the low-surface-brightness component instead that the value at the burst located at the edge of the galaxy (Paper II). Finally, two of these objects should certainly not be classified as BCDs but as dIrr, II Zw33 B (Walter et al. 1997) and UGC 4483 (van Zee, Skillman, & Salzer 1998b).
The physical reason for the segregation observed in B-band PSB between the different galaxy types is the presence in the BCDs of a recent star formation event that outshines the low-surface-brightness component. This event may be accompanied by blue optical colors and strong emission-lines (as in objects spectroscopically classified as HII galaxies). In dIrr and dE the recent star formation (if present) is comparatively less active and the PSB is consequently fainter and redder.
It is important to note that the fact that the segregation between BCDs and other types of dwarfs by PSB is better than the segregation by the color at the PSB is partially due to the different contribution of the red supergiants to the luminosity and color evolution of the burst (Doublier et al. 2001a).
Dwarf. One of the most important physical parameters driving the evolution of galaxies is the mass (see Brinchmann & Ellis 2000 and references therein). This is particularly important in the case of low-mass galaxies like the BCDs where it controls the formation of density waves or not. In this sense, the B-band luminosity cutoff imposed by Thuan & Martin (1981) was thought as a limit in the stellar mass of BCDs. However, the B-band luminosity is a very poor tracer of the stellar mass in a galaxy. Here we propose to use the K-band luminosity as a more reliable measure of the stellar mass in these galaxies (Gil de Paz et al. 2000a; Pérez-González et al. 2003a, b). In this regard it is worth mentioning that the assumption of different, plausible star formation histories within a galaxy may lead to changes in the mass-to-light ratio as high as a factor or 7 in the B-band but only a factor of 2 in K (Bell & de Jong 2001).
Unfortunately, the number of studies of BCDs in the near-infrared is still small and they are limited to a very few objects each (James 1994; Vanzi, Hunt, & Thuan 2002; Doublier, Caulet, & Comte 2001b; Noeske et al. 2003, submitted). In Figure 1c we have plotted the (B - K) and (B - R) colors for the 21 galaxies in our sample included in the 2MASS Second Incremental Release Extended Source Catalog (Jarrett et al. 2000). Dotted lines in this diagram represent the model predictions for a composite stellar population formed by a Z/5 metal-abundance burst with burst strength 1% (in mass) and age between 3.5 and 10 Myr. Thin, solid-lines correspond to the time evolution predicted by the models for the same burst with strength between 100% and 0.01%. The models used here are those developed by Gil de Paz et al. (2000a) and Perez-González et al. (2003a, b) which are based on the evolutionary synthesis models of Bruzual & Charlot (2003, unpublished). We have assumed that 15% of the Lyman continuum photons escape from the galaxy (or are absorbed by dust) before the ionization of the surrounding gas (see e.g. Gil de Paz et al. 2000a). This figure shows that there is a very good linear correlation between the integrated (B - R) and (B - K) color of these objects, which is also expected from the predictions of the models for burst strength values lower than 1% or (B - R) colors redder than ~ 0.5. The best fit to this correlation is
As we commented above this relation is applicable only when the (B - R) color is > 0.5 mag. However, as we will show in Section 8 virtually all the BCD galaxies with (B - R) color bluer than 0.5 mag have B-band absolute magnitude fainter than MB = -16.5, so they can undoubtedly classified as BCDs.
The average (B - K) color of the 21 sample galaxies within 2MASS is 2.82 ± 0.42. If we now apply this average (B - K) color to the limit in B-band luminosity imposed by Thuan & Martin (1981) we end up with an equivalent limit in K-band luminosity of MK > -21 mag (for H0 = 70 km s-1 Mpc-1). In Figure 1d we show the frequency histogram of K-band absolute magnitudes of the galaxies in our sample obtained applying the Equation 1. Noteworthy, from the 9 galaxies in our sample showing MB < -18.15 mag only 5 are brighter than MK = -21 mag. Within them IC10 is probably above this limit because of its uncertain galactic extinction correction. The other 4 galaxies should not be classified as BCD galaxies, II Zw 33, Mrk 7, Tol 1924-416, and Mrk 314. The galaxies showing MB < -18.15 mag but MK > -21 mag (KUG 0207-016A, Mrk 400, Haro 2, Pox 4) are probably relatively low-mass objects experiencing a very massive burst that makes their integrated colors bluer than the average. Individualized near-infrared observations are necessary to confirm this.
Summarizing, we propose that, in order to be classified as a BCD, a galaxy has to fulfill the following observational, quantitative criteria: (1) it has to be blue, µB, peak - µR, peak 1, (2) compact, µB, peak < 22 mag/arcsec2, and (3) dwarf, MK > -21 mag. As we have shown above these criteria recover most of the galaxies traditionally classified as BCDs and also allow to segregate the BCDs from other types of dwarf galaxies like dIrr or dE.